System for dynamic spectral correction of audio signals to compensate for ambient noise in the listener's environment

[0001] The present non-provisional patent application claims benefit to the earlier priority date of provisional patent applications, Ser. No. 61/494,227, filed Jun. 7, 2011, application Ser. No. 61/544,263 filed Oct. 6, 2011, application Ser. No. 61/632,402, filed Oct. 7, 2011, application Ser. No. 61/564,956 filed Nov. 30, 2011, application Ser. No. 61/527,757, filed Aug. 26, 2011, application Ser. No. 61/632,401 filed Oct. 7, 2011, application Ser. No. 61/564,965, filed Nov. 30, 2011, the disclosures of which are incorporated in their entirety herein by reference.

[0002] The present invention relates to the field of audio processing and more specifically to systems, devices, and methods to compensate for noise in a listener's environment.

[0003] There are many systems for the application of filtering to noise suppression of an audio signal. In most cases, these inventions relate to the removal of noise that is present in the audio signal from the origin or introduced into the signal through processing and transmission. Various forms of filtering may be applied which suppresses the noise signal in whole or in part. Generally, these systems have adverse impacts upon the quality of the original signal. However, these systems do not address noise in the environment of the listener, which cannot be filtered.

[0004] Conversely, systems for the suppression of noise in the listener's environment also exist. These systems generally use noise cancellation to remove the disrupting external signal by adding sound projected through headphones which has the effect of countering the sound waves produced by the noise. In this case, the noise is completely canceled and listener is generally unaware of the existence of the external noise, a result which can reduce the awareness of the listener to potential dangers in the environment.

[0005] In some prior art systems, dynamic volume compensation may be used to raise the volume of a source signal of interest over ambient background noise. However, these systems increase the gain in a spectrally uniform manner, raising the volume of all frequency components equally. This effect can distort the perception of music and speech due to the non-linear behavior of the human ear with respect to frequency and volume.

[0006] Microphones and mechanical systems (e.g., computer software) can measure dBSPL; a sound (e.g., 40 dBSPL) at particular frequency (e.g., 1 kHz) sounds just as loud as the sound (e.g., 40 dBSPL) at a different frequency (e.g., 4 kHz) to a microphone or mechanical system. However, our hearing can be affected by the mechanical construction of our outer ear and/or slow variation in sensitivity across the basilar membrane due to fluid damping of the incident waves in the cochlear fluid. The variable sensitivity of human hearing is reflected in the Fletcher-Munson equal loudness contours and the equal-loudness contours from ISO 226:3003 revision (Phons). The equations of the systems of the present invention utilize conversions from dBSPL to Phons and from Phons to dBSPL (incoming sounds levels are converted from dBSPL to Phons for use in the equations, then subsequently the Phons are converted to dBSPL for expression to speakers and headphones. Conversion from dBSPL to Phons and from Phons to dBSPL is in accordance with the Fletcher-Munson equal-loudness contours and the equal-loudness contours from ISO 226:3003 revision.

[0007] Since the human ear dynamically adjusts to sound intensity levels, the presence of background noise alters the threshold at which sounds begin to be perceived. As a result, ambient noise at a given frequency may make sounds at those frequencies that would otherwise be perceptible imperceptible in the presence of ambient noise. In order for the sound to be heard it must be amplified over the background noise. The volume of the ambient noise therefore represents a degree of hearing impairment or baseline threshold elevation over which the sound must be amplified to be perceived.

[0008] This effect varies according to the spectral composition of the noise, that is, spectral components that are sufficiently far from the spectral composition of the noise will remain perceptible. Consequently, using the total intensity of the background noise to raise the intensity of the source uniformly will overly amplify bands which are not affected, possibly raising the volume to damaging levels. In order to amplify only those components which need compensation, the gains to the source signal must vary by spectral band, according to the spectral composition of the noise.

[0009] Moreover, due to the nonlinear response of the human ear, using the spectral intensity of the background noise at a particular band as the gain for the source at that band will produce excessive amplification. In order to compute the correct gain a nonlinear psychoacoustic model must be used to compute an appropriate gain for each frequency. The intensity of the background noise as well as the source signal at a given frequency are inputs to this model, and the output is a desired gain for the source signal at that frequency.

[0010] The perceived loudness of sound can be modeled by a harmonic oscillator with spring constant that varies according to the mean power of vibration. This model is called the Earspring model. The Earspring equation 10,11 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration or S(P), identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor. Thus the resonant frequency of the Earspring varies with the amplitude of the <y²> term. Since <y²> is a function of y, this equation is nonlinear.

[0011] For a particular driving force F(t), which we will consider to be a sinusoid at a given frequency and amplitude, for which the amplitude is P in dBPhons, a solution to the earspring equation y(t) can be found, for a particular set of boundary conditions, which is the steady state response to the forcing function. Transforming into the frequency domain we can obtain the mean power of vibration <y²>=½|Y|², which can be rewritten as S(P).

[0012] A possible set of boundary parameters comprises (i) a Sones ratio between the hearing threshold level and the reference level of 40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz; This equation can be solved for numerically though in practice it is more efficient to use a computational model to estimate the solution.

[0013] We can derive an equation for S(P),

[0014] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10}$
where S(P) is the perceived loudness in Sones which the listener experiences, and P is the sound intensity impinging on the ear in dBPhons. Where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency for a ˜40 dBSPL amplitude driving force. The constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions. Note that this formula is independent of frequency although the sound intensity P of any given signal may vary as a function of frequency.

[0015] The present invention features systems for dynamically adjusting audio signals by applying a gain to the signal in a spectrally varying manner to compensate for a noise source in the environment of the listener, such that the sound is perceived by the listener to be unchanged in loudness and spectral composition. The system obtains a threshold elevation for each frequency component by analyzing the spectral composition of the ambient noise. This threshold elevation is then used by a psychoacoustic model of hearing to determine an appropriate gain adjustment for the corresponding frequency component of the source signal which will make that component perceived by the human ear to be just as loud as if the noise were not present. After applying the gains to the source signal, the resulting signal is output to the speaker. The system allows a listener to hear without distortion, over ambient noise, by applying a gain to the source that varies according to the spectral composition of the noise, rather than cancelling the noise, or applying a uniform volume adjustment to the source. The perceived spectral composition of the source is thus adjusted without the removal of the noise signal. Systems may be incorporated into apparatuses including but not limited to mobile phones and music players.

[0016] Any feature or combination of features described herein are included within the scope of the present invention provided that the features included in any such combination are not mutually inconsistent as will be apparent from the context, this specification, and the knowledge of one of ordinary skill in the art. Additional advantages and aspects of the present invention are apparent in the following detailed description.

The Following Information Regarding Terms is Non-Limiting and Exemplary in Nature for the Purpose of Understanding the Spirit of the Invention

• • [0017] 1. Amplifier: This component can be any amplifier or related device capable of applying gains to an audio signal.
  • [0018] 2. Audio Codec: Audio Encoder/Decoder converts an audio signal to and from its encoded format either for storing in a file or transmission over a network
  • [0019] 3. A/D Converter: Converts audio signals from analog to digital format. In some embodiments, the term A/D Converter is used interchangeably with a means for converting an audio signal from analog to digital form. Analog-to-digital and digital-to-analog converters can be assumed to exist at interface points between analog elements and digital elements of any embodiment. Said means may be electronic, non-electronic or only partially electronic. Said converters, also commonly referred to as Analog to Digital Converters or ADCS, are well known within the art and would be understood by one skilled in the same. Said converters could include but are not limited to direct-conversion ADCs, successive approximation ADCs, a ramp-compare ADC, Wilkinson ADCs, integrating ADCs, Delta-encoded ADCs, pipeline ADCs, time interleaved ADCs, or ADCs with intermediate FM Stage
  • [0020] 4. Audio Device: As used herein, an audio device is any audio device capable of extracting, storing or receiving an audio signal. In some embodiments, the term audio device is used interchangeably with a “means for extracting and/or identifying an audio signal of interest and/or ambient noise signal.” Said means may be, for example, a microphone, receiver, or storage media containing a signal of interest in an audio file and code for reading the signal from an audio file.
  • [0021] 5. Audiogram: One form of defining a person's hearing loss which plots thresholds of hearing relative to a standardized curve that represents ‘normal’ hearing, in dB(HL). An audiogram can be obtained using a behavioral hearing test called Audiometry. The “Tone” test involves presenting different tones at a specific frequency (pitch) and intensity (loudness) and determining thresholds of hearing for each tone where the thresholds correspond to how loud the tone must be presented in order for the person to perceive the tone sound.
  • [0022] 6. Audio Jack: An input by which an audio signal can be received. i.e. from a microphone.
  • [0023] 7. Background signal: The background signal, as used herein, refers to a portion of the second audio signal or the ambient noise signal at the same frequency component as that of the source signal. The source signal and the background signal are paired.
  • [0024] 8. Calibration equation: Calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed. It is represented by: P_dbSPL(F)=P_dBFS(F)+(P_dBSPL0(F)−P_dBFS0(F)) wherein P_dBSPL0 is nominally a value between 65 and 83, and wherein P_dBFS0 is nominally a value between −20 and −12;
    • [0025] where P=Phons amplitude of a source signal of interest
    • [0026] F=Frequency
    • [0027] dBSPL=decibels according to sound pressure
    • [0028] dBFS=decibels relative to full scale
  • [0029] 9. Bark frequency: The bark scale is a psychoacoustic scale for subjective measurements of loudness. The scale is broken down into critical bands of hearing. Critical bands are known frequency bandwidths that correspond to limitations of the human ear to perceive sounds with frequency differences smaller than the critical bandwidths.
  • [0030] 10. D/A Converter: Converts audio signals from digital to analog format. In some embodiments, the term D/A Converter is used interchangeably with a means for converting an audio signal from digital to analog form. Analog-to-digital and digital-to-analog converters can be assumed to exist at interface points between analog elements and digital elements of any embodiment. A D/A converter, or DAC, is capable of converting a digital, usually binary signal code to an analog signal (current voltage or electric charge). DACs may include but are not limited to pulse width modulators, oversampling or interpolating DACs, binary weighted DACs, R-2R Ladder DACs, Successive approximation or cyclic DACs, thermometer coded DACs, and hybrid DACs.
  • [0031] 11. Digital Signal Processor (“DSP”) Chip: A specialized microprocessor with an architecture for the fast operational needs of digital signal processing.
  • [0032] 12. FFT Co-Processor: A specialized microprocessor designed for the purpose of rapidly computing Fourier transforms using a algorithm for computing Fourier transforms commonly known as the ‘Fast Fourier Transform’ (FFT).
  • [0033] 13. Field Programmable Gate Array (“FPGA”): a reprogrammable electronic chip.
  • [0034] 14. Filter banks: An array of band-pass filters that separate an input signal into multiple components, or as used herein, frequency components, each frequency component carrying a single frequency subband of the original signal.
  • [0035] 15. FIR Filter: A filter with an impulse response (or response to any finite length input) of finite duration, because it settles to zero in finite time. This is in contrast to infinite impulse response (IIR) filters, which have internal feedback and may continue to respond indefinitely (usually decaying). FIR filters can be discrete-time or continuous-time, and digital or analog.
  • [0036] 16. Fourier transforms: A mathematical operation that computes the frequency content of an audio signal, taking a given audio signal as input and outputting magnitudes and phases across the frequency spectrum.
  • [0037] 17. Frequency Analysis Model: This model can use a variety of known methods to divide the audio signal of interest into a plurality of frequency components. Such methods may include, but are not limited to methods such as filter banks, Fourier transforms, wavelet transforms or other signal processing techniques. In some embodiments, the term Frequency Analysis Model is used interchangeably with a “means for dividing an audio signal into a plurality of frequency components. Said means may be, for example, a digital filter bank, analog filter bank, FFT Co-processor, or code for computing a Fourier transform analysis to an audio signal on computer-readable media.
  • [0038] 18. Frequency component and Frequency Component Spectrum: A frequency component, as used herein, identifies a portion of the frequency range of an audio signal. A frequency component can may comprises a particular, individual frequency, frequency channels, or frequency bands. A frequency component spectrum is a plurality of frequency components.
  • [0039] 19. Frequency Synthesis Module: A module which the present invention can use to reconstitute the various frequency components of a particular audio signal of interest. The frequency synthesis Module may perform summation of the various frequency components of the signal of interest, or may perform an inverse Fourier transform, or other transform, after said frequency components have been adjusted, to create a new signal or waveform, depending on whether the Frequency analysis module was a filter bank or a fourier transform, or other transform. “Sound Systems: Design and Optimization: Modern Techniques and Tools for Sound System Design and Alignment”, McCarthy, Bob (2010) (Focal Press) Second Edition. In some embodiments, the term Frequency Synthesis Module is used interchangeably with a “means for reconstituting an audio signal”. Said means may be, for example, a summer block, a frequency synthesis module, an FFT-Co processor, an inverse FFT transform, or code for implementing an inverse FFT transform on computer-readable media.
  • [0040] 20. Gain Block component—This component applies the gains to each corresponding frequency component or band. In some embodiments, the term gain block component is used interchangeably with a “means for applying gains to a audio signal.” Said means may be, for example, an amplifier, or gain block component.
  • [0041] 21. Infinite Impulse Response (“IIR”) Filter: A signal processing filter with an impulse response function that is non-zero over an infinite length of time. May be implemented as either digital or analog IIR filters. IIR filters use fewer computing resources because they use fewer taps.
  • [0042] 22. Input Device: An device configured to enable a user to set a parameter b of a personalized sound adjustment model. In some embodiments, the term input device is used interchangeably with “a means for enabling a user to set a parameter b of a personalized sound adjustment model”. Said means may be, for example, a physical input device such as a user adjustable input, a knob, a slider, or any combination thereof. Said means may alternatively be an electronic means or a combination of physical hardware and electronics such as graphical user interface. Said means may further be a touch screen configured to accept a user's choice of parameter b. Additional means would be recognized and understood by one skilled in the relevant art.
  • [0043] 23. Inverse Fourier Transform: A mathematical operation which computes a time domain signal corresponding to a given frequency spectrum.
  • [0044] 24. kHz frequency to critical band center frequency converter: The critical bands of hearing are natural bands in the audio frequency spectrum in the human ear. These bands are characterized by the masking of tones by other tones in the same critical band. There are several experimentally derived frequency scales that attempt to approximate the critical bands including the Bark scale—a psychoacoustic model proposed by Eberhard Zwicker in 1961, and named after Heinrich Barkhausen who proposed the first subjective measurements of loudness. One example of the equation for converting from kHz to critical band frequency is the conversion to Bark frequency:
    CBR(F_kHz)=26.811(1+1.960/F_kHz)−0.53
  • [0045]  In some embodiments, the term kHz frequency to critical band center frequency converter is used interchangeably with a “means for converting a frequency component or series of frequency components to critical band frequency. Said mean may be, for example, a summer.
  • [0046] 25. Linear slope equation: This equation is part of the Personalized sound adjustment model, and is a novel mathematical equation, and one potential embodiment by which threshold elevations are generated for vTuning. In some embodiments, the term Linear Slope equation is used interchangeably with a “means for estimating threshold elevations specific to a particular person.” Said means may be, for example, the Linear Slope equation, an audiogram, or a pattern library.
  • [0047] 26. Lookup Table: A tabular list of parameters indexed by some given variable, such as signal power. Used by the model to select a parameter or set of parameters depending on the current real-time value of a variable. In some embodiments, the term Lookup Table is used interchangeably with a “means for determining threshold elevations by using a lookup table.” Said means may be, for example, computer-readable media code for retrieving a threshold elevation value from a memory device using a user adjustable parameter b and frequency as indexes.
  • [0048] 27. Microphone: The microphone can be any microphone or the like capable of receiving an audio signal.
  • [0049] 28. Parallel Compression: applying a linear gain to a signal which amplifies softer sounds and subsequently adding this amplified sound back in to the original signal. Also known as New York compression. In some embodiments, the term parallel compression is used interchangeably with a “means for computing and/or applying corrective gains to an audio signal where said gains correspond to threshold elevations, using a combination of parallel compression and known psychoacoustic models” or a “means for using parallel compression to approximate a desired non-linear dynamic range compression curve.” Said means can for example, be any known psychoacoustic model such as the model disclosed herein. Said means may for example, be a model where a compression curve defines the output loudness in decibels versus the input loudness in decibels of a signal of interest, wherein the compression is implemented by applying a gain G to the signal of interest according to the following equation:
    G=g0+(P−T)*(1/R−1)
  • [0050]  where G is the gain in dB to be applied to a frequency component of the signal of interest, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of linear compression, whereupon the linearly compressed signal is added together with the original signal of interest, producing a parallel compressed output signal, where the overall amplitude of the input versus the output signal defines a parallel compression curve, where g0, T, and R are thus parameters that vary the shape of the parallel compression curve, and P is the input loudness of the signal of interest. In some embodiments, the term parallel compression is used interchangeably with a “means for applying parallel compression to a signal of interest.” Said means may be, for example, a parallel compression chip or combining a compressed signal with an original signal. In some embodiments, parallel compression may be a “means for applying linear compression to a signal.” Said means may for example be an amplifier. In some embodiments, the term parallel compression is used interchangeably with a “means for controlling the shape of a parallel compression curve.” Said means may be, for example threshold elevations, makeup gains, and compression ratios. In some embodiments, the term parallel compression is used interchangeably with a “means for finding the parameters which best fit a parallel compression curve against a psychoacoustic model”. Said means may be, for example, linear regression, least squares fit, and function approximation. In some embodiments, the term parallel compression is used interchangeably with a “means for storing parameters”. Said means may be, for example, use of a lookup table, flash memory, hard disk, and long-term memory. In some embodiments, the term parallel compression is used interchangeably with a “means for retrieving the parameters to be applied to a parallel compression module.” Said means may be, for example, a lookup table, polynomials, and a parameter estimation module.
  • [0051] 29. Parameter: The parameter is a single mathematical parameter value that when input to the Personalized Sound adjustment, generates may generate a spectrum of threshold elevations.
  • [0052] 30. Pattern Library: A numerical library composed of points and values which measure threshold elevations. In some embodiments, the term Pattern Library is used interchangeably with a “means for creating a pattern library for storing threshold elevations.” Said means may be, for example, obtaining sample threshold elevation patterns for varying frequency for various typical users (users indexed by b) or computer-readable media code for storing a set of patterns in memory indexed by a user adjustable parameter b and a frequency.
  • [0053] 31. Personalized sound adjustment model: A novel mathematical model which enhances sound through a process of non-linear compression. Bark frequency is one of several types of measures of the natural bandwidth of human hearing referred to as critical bandwidth. Critical bands are defined by the distances in frequency in which masking of tones occurs in human ear, roughly correspond to linear position on the cochlea. Thus, the present invention utilizes an equation that, using a single parameter, chosen by a user, generates a spectrum of threshold elevations that generally fall along a straight line when plotted on models using dBHL against Bark frequency where said threshold elevations can be used to adjust a sound signal for persons with hearing loss or persons who simply wish to enhance the audio signal to their preference. In some embodiments, the term personalized sound adjustment model is used interchangeably with “a means for executing a personalized sound adjustment model”. Said means may be computing the model on a microprocessor, a digital signal processor, or using code for executing the model on computer-readable media. In some embodiments, the term personalized sound adjustment model is used interchangeably with a “means for determining threshold elevations from ambient noise.” Said means may be, for example, dividing an ambient noise signal into frequency components, estimating the ambient noise power of a frequency component, and determining threshold elevations corresponding to a frequency component.
  • [0054] 32. Power Estimation Model/Block: estimating the power value of an audio signal. In some embodiments, the term Power Estimation Block/Model is used interchangeably with a “means for estimating a power value of an audio signal.” Audio power estimation can be conducted using a variety of known techniques which would be understood by one skilled in the art. For example, some use a Minimum Statistics approach which is based on tracking minima of a short term power estimate of the audio signal in bands, over time. If a Fourier transform of the signal is used the power estimate is the square of the magnitude of the frequency component.
  • [0055] 33. Psychoacoustic model: Any appropriate psychoacoustic models can be used by the present invention to compute gains needed to amplify sound to overcome the effects of ambient noise on sound perception or gains needed to account for preferred threshold elevations computed for users' of normal hearing. One example of a known psychoacoustic model is as follows:

[0056] $\frac{\sqrt[\alpha ]{P_{\mathrm{SIG}}^{\alpha }+P_{\mathrm{NOISE}}^{\alpha }-P_{\mathrm{THRQ}}^{\alpha }}}{P_{\mathrm{SIG}}}$

• • [0057] 34. where G is the gain ratio, PSIG is the signal intensity at a frequency in units of power, PNOISE is the signal intensity of the background noise, PTHRQ is the absolute threshold of hearing, and α=0.2 is a constant. In some embodiments, the term psychoacoustic model is used interchangeably with a “means for generating or computing corrective gains corresponding to predetermined threshold elevations”. Said means may be, for example, additional known psychoacoustic models, perceptual models, computational models, and models utilizing parallel compression fit against known psychoacoustic models.
  • [0058] 35. RF Transceiver: Radio Frequency Transmitter/Receiver—this device interfaces directly with the antenna and can modulate an audio signal with its radio carrier frequency for transmission as well as demodulate a received radio frequency signal
  • [0059] 36. Source signal: The Source signal, as used herein refers to a portion of the first audio signal or the audio signal of interest at a given frequency component.
  • [0060] 37. Speaker: The speaker can be any speaker or the like capable of receiving and projecting a sound signal adjusted or corrected by the present invention.
  • [0061] 38. Storage media: A computer readable media used for storing data, such as a hard disk, RAM, or flash memory
  • [0062] 39. Summer block: Summation occurs when two or more audio signals are combined together to create a new wave form. Summation is the combining of two audio signals at the same frequency—not to be confused with mixing, which involves the combining of audio signals at different frequencies. Electrical summation occurs inside an electrical circuit. In some embodiments, the term summer block is used interchangeably with a “means for combining an audio signal with another audio signal”. Said means may be, for example, a summer.
  • [0063] 40. Threshold elevation: Threshold elevations correlate to the minimum sound pressures required to perceive sounds at various frequencies.
    Additional Terms:
  • [0064] 1. Conductor Equation (Perceptual Model Component): This is a novel equation which computes the gain necessary to amplify sound sufficiently to make the subjectively perceived sound appear to be as loud, to a hearing impaired subject, as if the subject had no hearing loss. In some embodiments, the term Conductor equation is used interchangeably with a means for determining the gain necessary to make a sound perceived to be as loud as if a threshold elevation were not present.
  • [0065] 2. dBSPL: Decibels sound power level—a unit of measure of sound intensity at the ear
  • [0066] 3. dBSPL to dBPhon Conversion equation: This equation converts sound pressure level to Phons in decibels:

[0067] $P_{\mathrm{dBPhon}}\left(F\right)=\frac{P_{\mathrm{dBSPL}}\left(F\right)-\mathrm{ATH}\left(F\right)}{1+\mathrm{kATH}\left(F\right)}$

• • [0068]  where
    • [0069] F=Frequency
    • [0070] ATH=Absolute Threshold of Hearing
    • [0071] P=Phons amplitude of a source signal of interest
    • [0072] dBSPL=decibels according to sound pressure
  • [0073] 4. dBFS to dBSPL Converter: This equation converts the audio signal power from audio power in decibels to audio sound pressure level in decibels
  • [0074] 5. Ear Spring Equation/Crescendo (Psychoacoustic Model): This is a novel equation which computes the apparent subjective loudness for a normal subject, of a sound in Sones as a function of the input sound intensity of the sound in Phons. In some embodiments, the term “Ear Spring Equation/Crescendo” is used interchangeably with a “means for computing the loudness in Sones from the intensity in Phons at a person ear.” Said means may be the Earspring Equation and may also be any known psychoacoustic models.
  • [0075] 6. Gain Equation: This equation converts the gain in Phons to a gain in sound pressure level
    ΔP_dBSPL(F)=ΔP_dBPhon(F)(1+kATH(F))
  • [0076]  where:
    • [0077] F=Frequency
    • [0078] ATH=Absolute Threshold of Hearing
    • [0079] P=Phons amplitude of a source signal of interest
    • [0080] dBSPL=decibels according to sound pressure
    • [0081] K=− 1/120 when ATH(F)>0
      • [0082] =+ 1/240 when ATH(F)<0
  • [0083] 7. Output Device: A device configured to output an audio signal. In some embodiments, the term Output Device is used interchangeably with Audio Device. In some embodiments, the term Output Device is used interchangeably with a “mean for outputting an audio signal.” Said means may be, for example, a speaker, a transmitter, computer-readable media code for transmitting an audio signal or audio file and computer-readable media code for writing an audio signal to an audio file.
  • [0084] 8. Phons: Phons are objective units of sound intensity impinging upon the ear. Sound power may be converted from decibels (dBFS) to Phons using a conversion first to dBSPL then to Phons
  • [0085] 9. Psychoacoustic model (Crescendo): This is a novel equation derived from a complex combination of physics and mathematical equations, which computes gains needed to compensate for ambient noise using two major input values, a source signal of interest, and an ambient noise signal. In some embodiments, the term psychoacoustic model is used interchangeably with a “means for determining the gains to be applied to a signal of interest.” Said means may be, for example, a known psychoacoustic model or Crescendo.
  • [0086] 10. Sones: Sones are subjective units of perceived sound loudness.
  • [0087] 11. Square Summer Equation: This equation squares the audio signal and adds the signal up N times, whereby the sum of the signal at the end block, divided by N, generates a mean power value for the audio signal. In some embodiments, this term is used interchangeably with a “means for extracting power values of an ambient noise signal.”

[0103] As shown in FIG. 1, in some embodiments, the present invention can comprise an apparatus 100, effective in enhancing an audio signal, by applying corrective gains to a plurality of frequency components which constitute a first audio signal, wherein the value of the gains are computed using a nonlinear psychoacoustic model of loudness as a function of sound intensity, as subjectively experienced by a normal human, where this psychoacoustic model takes as inputs the sound intensity of the audio signal P as well as the background noise P₀, wherein corrective gains are generated for each channel by solving for a correction gain ΔP for the signal which will suffice to raise the volume of each channel so that the user will perceive it above background noise, by the loudness amount equivalent to how it would be perceived in a noiseless environment

• • [0104] the apparatus 100 comprising:
  • [0105] (a) a first audio source 02, configured to extract the first audio signal where the first audio signal is a broadband audio signal of interest, the first audio source operatively connected to a microprocessor 01;
  • [0106] (b) a second audio source 03, configured for extracting ambient broadband audio in the environment, the second audio source operatively connected to the microprocessor 01;
  • [0107] (c) the microprocessor 01, operatively connected to the first audio source 02, the second audio source 03, a Speaker 20, and an Amplifier 19; the microprocessor 01: configured to execute applications for:
    • [0108] i. a first Analog to Digital converter 04;
    • [0109] ii. a second Analog to Digital converter 05;
    • [0110] iii. a first Frequency Analysis Module 06;
    • [0111] iv. a second Frequency Analysis Module 07;
    • [0112] v. a first Power Estimator Block 08;
    • [0113] vi. a second Power Estimator Block 09;
    • [0114] vii. a first dBFS to dBSPL Converter 10;
    • [0115] viii. a second dBFS to dBSPL Converter 11;
    • [0116] ix. a first dBSPL to dB Phon Converter 12;
    • [0117] x. a second dBSPL to dB Phon Converter 13;
    • [0118] xi. a Psychoacoustic model 22;
    • [0119] xii. a Gain Block component 17;
    • [0120] xiii. a Frequency Synthesis Module 18; and
    • [0121] xiv. a Digital to Analog Converter 19;
      whereupon (i) the first audio source 02 extracts the first audio signal x(t), where the first audio signal is then transmitted to the first Analog to Digital Converter 04 thereby converting the first audio signal into digital format; and simultaneously, (ii) the second audio source 03 extracts a second audio signal x₀(t), where said second audio signal is then transmitted to the second Analog to Digital Converter 05 thereby converting the second audio signal into digital format;
      whereupon the first audio signal and the second audio signal are fed, respectively, through the First Frequency Analysis Module 06 and Second Frequency Analysis Module 07 breaking down each respective audio signal into separate components;
      whereupon the apparatus 100 identifies the source signal and the background signal corresponding to a frequency 50 and said signals are thereafter transmitted to the microprocessor 01;
      whereupon both the source signal and the background signal are fed, respectively, through the first Power Estimator 08 and the Second Power Estimator 09, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS 52;
      whereupon (i) the microprocessor 01 completes a calibration offset such that a first dBFS to dBSPL Converter 10 converts the power value of the source signal from dBFS to dBSPL, and (ii) the microprocessor 01 completes a calibration offset such that a second dBFS to dBSPL Converter 11 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f);
      whereupon (i) a first dBSPL-dB Phon Converter 12 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 13 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f);
      whereupon the microprocessor 01 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the psychoacoustic model 22 whereupon the microprocessor 01 calculates ΔP(f) in dBPhons
      whereupon the microprocessor 01 calculates a correction gain of power to the source signal G(f), at the frequency, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 17 which utilizes the Gain equation;
      whereupon the apparatus 100 reconstitutes the first audio signal with the corrective gain to the source signal corresponding to the frequency, comprising:
  • [0122] (i) multiplying the source signal at the frequency according to the corresponding corrective gain ΔP(f);
  • [0123] (ii) using the Frequency Synthesis Module 18, to reconstitute the resulting audio signal
    whereupon the corrected audio signal is fed through the Digital to Analog Converter 19, resulting in analog signal x_out(t);

[0124] As shown in FIG. 2, in some embodiments, the psychoacoustic model involves analytically computing, using the Earspring equation the value in Sones of the input signal at a given frequency S(P) 114, along with the estimated hearing loss due to background noise in Sones at that frequency S(P₀) 115, and applying these values to the conductor equation to derive a gain in Phons for the signal that will correct for the perceived hearing loss, the Conductor Equation 116 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P

[0125] As shown in FIG. 3, in some embodiments, the psychoacoustic model can use a numerical estimate of these equations, combining the conductor equation and the Earspring equation solutions into one equation. In this embodiment, a function approximator 124, such as a quadratic estimate, linear regression with quadratic feature, neural network, or other computational model, can be fit against the solutions to the Earspring equation, and the parameters of the model then stored in memory in a lookup table 123. During operation, these parameters can be retrieved by the microprocessor from a lookup table, and used to compute an approximate value for ΔP for any given P and P₀.

[0126] As shown in FIG. 4, in some embodiments, an alternative means of effectuating the gains demanded by the psychoacoustic model is to harness an existing sound processing technique known as parallel compression. Parallel compression involves compressing a copy of an original signal using linear compression 112, and subsequently adding this compressed signal back into the original signal. This results in a non-linear compression curve that amplifies low sounds below a certain threshold. The shape of the compression curve is controlled by a number of parameters including the threshold (T), compression ratio (R), and makeup gain (M), and applies gains that vary depending on the power of the signal of interest 106. By varying these parameters, the parallel compression curves can be made to approximate the compression curve that would otherwise be produced by applying the psychoacoustic model. This is done by fitting the parameters against the desired characteristics of the psychoacoustic model using a least-squares error fit. These parameters are then stored in memory in a lookup table 111, or computed using a formula and may be retrieved in real time to dynamically update the parallel compression block 112. Thus, these parameters must be updated at each time step account for the changing noise. In this case, the noise power 107 is used to determine the parameters of the compression gain. Typically the T parameter is chosen to be the threshold elevation (or noise power) minus 20 dB. From that point makeup gain and compression ratio are found using standard curves that have been derived by fitting against the psychoacoustic model. Plots of the parameters used for varying threshold elevations are shown in FIG. 5 and FIG. 6. Here we can see that the desired compression ratio has an “upside down U” shape for varying threshold elevations.

[0127] As shown in FIG. 7, in some embodiments, the present invention comprises, an apparatus 200, effective in enhancing an audio signal, by applying corrective gains to a plurality of frequency components which constitute a first audio signal, wherein the gains are applied using a parallel compression technique, where the parameters of the parallel compression are determined by fitting the compression curves against a non-linear psychoacoustic model, where the parameters depend upon the power of a background noise signal P₀, and where the compression is applied to the first audio signal, the apparatus 200 comprising:

• • [0128] a) a first audio source 202, configured to extract the first audio signal where the first audio signal is a broadband audio signal of interest, the first audio source operatively connected to a microprocessor 201;
  • [0129] b) a second audio source 203, configured for extracting ambient broadband audio in the environment, the second audio source operatively connected to the microprocessor 201;
  • [0130] c) the microprocessor 201, operatively connected to the first audio source 202, the second audio source 203, a Speaker 220, and an Amplifier 219; the microprocessor 201 configured to execute applications for:
    • [0131] i. a first Analog to Digital converter 204;
    • [0132] ii. a second Analog to Digital converter 205;
    • [0133] iii. a first Frequency Analysis Module 206;
    • [0134] iv. a second Frequency Analysis Module 207;
    • [0135] v. a first Power Estimator Block 208;
    • [0136] vi. a second Power Estimator Block 209;
    • [0137] vii. a first dBFS to dBSPL Converter 210;
    • [0138] viii. a second dBFS to dBSPL Converter 211;
    • [0139] ix. a first dBSPL to dB Phon Converter 212;
    • [0140] x. a second dBSPL to dB Phon Converter 213;
    • [0141] xi. a lookup table 215;
    • [0142] xii. a parallel compression module 214;
    • [0143] xiii. a Frequency Synthesis Module 218; and
    • [0144] xiv. a Digital to Analog Converter 219;
      whereupon (i) the first audio source 202 extracts the first audio signal x(t), where the first audio signal is then transmitted to the first Analog to Digital Converter 204 thereby converting the first audio signal into digital format; and simultaneously, (ii) the second audio source 203 extracts a second audio signal x₀(t), where said second audio signal is then transmitted to the second Analog to Digital Converter 205 thereby converting the second audio signal into digital format;
      whereupon the first audio signal and the second audio signal are fed, respectively, through the first First Frequency Analysis Module 206 and Second Frequency Analysis Module 207 breaking down each respective audio signal into separate components;
      whereupon, for each frequency component, the apparatus 200 identifies the source signal and the background signal corresponding to a frequency 250 and said signals are thereafter transmitted to the microprocessor 201;
      whereupon, for each frequency component, both the source signal and the background signal are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS 252;
      whereupon, for each frequency component, (i) the microprocessor 201 completes a calibration offset such that a first dBFS to dBSPL Converter 210 converts the power value of the source signal from dBFS to dBSPL, and (ii) the microprocessor 201 completes a calibration offset such that a second dBFS to dBSPL Converter 211 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f);
      whereupon, for each frequency component, (i) a first dBSPL-dB Phon Converter 212 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 213 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f);
      whereupon, for each frequency component, the microprocessor 201 uses the power value of the background signal to determine parameters T, R, and M for the parallel compression module 215,
      whereupon, for each frequency component, the parallel compression 215 module amplifies the source signal, by the compressive gains determined by the parameters,
      whereupon the parallel compression module sums the amplified source signal with the original source signal, to produce a corrected source signal,
      whereupon the apparatus 200 reconstitutes the plurality of frequency components of the corrected source signal using the Frequency Synthesis Module 218, to produce a corrected output audio signal;
      whereupon the corrected audio signal is fed through the Digital to Analog Converter 19, resulting in analog signal x_out(t);

[0145] As shown in FIG. 8, in some embodiments, the present invention comprises digital filter banks 306, 307 composed of IIR filters, in which case the frequency synthesis module is a summer block 318 that adds together the separate channels produced by the digital filer.

[0146] As shown in FIG. 9, in some embodiments, a combination of different types of frequency analysis modules and psychoacoustic models may be used, for instance combining digital filter banks and a computational model for the psychoacoustic model.

[0147] As shown in FIG. 10, in some embodiments, Fourier transforms may be used, using FFT co-processors to speed computation, in which case the frequency synthesis module is an FFT co-processor that performs an inverse Fourier transform on the signal.

[0148] In some embodiments, all or part of the present invention could be implemented using a parallel processing architecture such that some or all of the set of computations from step 5 to 10 above may be computed simultaneously using multiple processing units.

Computer Readable Media

[0149] In some embodiments, as explained by FIG. 11-14, the present invention can exist in major part on a computer medium, 670 storing a set of instructions executable by one or more processors, where the computer-readable medium 670 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of channels which constitute a first audio signal, wherein corrective gains are generated for each channel by solving for a correction value of ΔP for a source signal, wherein said source signal is a channel of the audio signal corresponding to a particular frequency, where ΔP is derived from a novel Conductor Equation 616,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,
where S(P)=a Sones equation from an EarSpring equation solution 614, as a function of Phons Pressure, P=a Phons amplitude of a source signal of interest, ΔP=the gain needed to make a source signal appear at proper loudness for impaired hearing, and P₀=an elevated threshold in Phons of impaired hearing, the computer-readable medium comprising:

• • [0150] (a) code for solving for S(P) of the Conductor Equation which comprises:
    • [0151] (i) code for extracting a power value of the source signal at a frequency, in dBFS with a first power estimator block, comprising:
      • [0152] (A) extracting the first audio signal x(t) 602, said first audio signal comprising a plurality of channels corresponding to varying frequencies, where said first audio signal is then transmitted to an Analog to Digital Converter thereby converting the signal into digital format 604;
      • [0153] (B) feeding the first audio signal through a first digital filter bank 606 where the first audio signal is broken down into separate channels corresponding to varying Frequencies f;
      • [0154] (C) identifying a source signal, wherein the source signal is a particular channel of the first audio signal corresponding to the frequency 650;
      • [0155] (D) feeding the source signal through a squared summer equation 651 whereby the source signal is squared and added up N times, whereby the sum of the signal at the end block divided by N generates a mean power value for the source signal, whereby said power value is then fed into a dB Converter to generate a power value for the source in dBFS 652, P_dBFS(f);
    • [0156] (ii) code for calculating a calibration offset to convert the power value of the source signal P_dBFS(f) in dBFS to a power value P_dBSPL(f) in dBSPL 610 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12;
    • [0157] (iii) code for converting the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation 612, producing P(f);
    • [0158] (iv) code for applying the power value of the source signal P(f), now in dB Phons, to the EarSpring equation 614 in order to obtain the power value of the source signal in Sones, or S(P(f));
  • [0159] (b) code for solving for S(P₀) of the Conductor Equation which comprises:
    • [0160] (i) code for extracting a power value P₀(f) of the background signal at the frequency, in dBFS with a second Power Estimator Block, comprising:
      • [0161] (A) extracting a second audio signal x_o(t) 603, said second audio signal comprising a plurality of channels corresponding to varying frequencies, where said second audio signal is then transmitted to a second analog to digital Converter 605 thereby converting the signal into digital format;
      • [0162] (B) feeding the second audio signal through a second Digital Filter Bank 607 where the second audio signal is broken down into separate channels X₀(f) corresponding to varying frequencies;
      • [0163] (C) identifying the background signal 650, wherein the background signal is a particular channel X₀(f) of the audio signal corresponding to the frequency;
      • [0164] (D) feeding the background signal through a squared summer equation 651 whereby the background signal is squared and added up N times, whereby the sum of the signal at the end block divided by N generates a mean power value for the signal, whereby said power value is then fed into a dB Converter to generate a power value for the background signal in dBFS 652, producing P_0—dBFS(f);
    • [0165] (ii) code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dBSPL(f) in dBSPL using the calibration equation 611 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12;
    • [0166] (iii) code for converting the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation 613, producing P₀(f);
    • [0167] (iv) code for applying the power value of the background signal, now in dB Phons, to the EarSpring equation 615 in order to obtain the power value of the background signal in Sones, or S(P₀);
  • [0168] (c) code for using S(P(f)) and S(P₀(f)) to solve for ΔP(f) using the Conductor equation 616 which results in the Conductor equation 616 generating a ΔP(f) in dBPhons;
  • [0169] (d) code for calculating a correction gain of power G(f), for a frequency f, in dBSPL based on the value of ΔP(f) in dB Phons, using a Gain equation 617;
  • [0170] (e) code for reconstituting the first audio signal with the corrective gain to the source signal at the frequency, comprising
    • [0171] (i) feeding the correction gain to an amplifier wherein the amplifier multiplies the source signal at the frequency by the corrective gain ΔP(f) 653;
    • [0172] (ii) combining, using a Summer Block 618, the channels of the corrected audio signal to reconstitute the corrected output x_out(t);
  • [0173] (f) code for feeding the first audio signal through a Digital to Analog Converter 619;
  • [0174] (g) code for transmitting the first audio signal, now in analog format, to the Amplifier 620; and
  • [0175] (h) code for causing the Amplifier 620 to output the audio signal, now containing the corrective gain at the frequency, to a Speaker 620 where the Speaker 620 outputs the audio signal with the corrective gain at the frequency;
    wherein the EarSpring equation is [d²/dt²+2β(d/dt)+k(1+γγ<y²>))]y(t)=ηF(t), where t=time, y(t)=amplitude of vibration, F(t)=driving force, identified as the Phons amplitude in the following equations, <y²>=mean power of vibration or S(P), identified as the Sones power in the following equations, β=damping constant, k=spring constant, and γ=coefficient of power dependence of spring constant, and η=the scale factor;
    wherein the set of boundary parameters comprises (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz;
    wherein the dBSPL to dBPhon Conversion equation 608 is:

[0176] $P_{\mathrm{dBPhon}}\left(F\right)=\frac{P_{\mathrm{dBSPL}}\left(F\right)-\mathrm{ATH}\left(F\right)}{1+\mathrm{kATH}\left(F\right)}$
wherein the Gain equation 613 is:
ΔP_dBSPL(F)=ΔP_dBPhon(F)(1+kATH(F))
where

• • [0177] F=frequency
  • [0178] ATH(F)=Absolute Threshold of Hearing
  • [0179] K=− 1/120 when ATH(F)>0
    • [0180] =+ 1/240 when ATH(F)<0

[0181] In some embodiments, the absolute threshold of hearing may be computed in one example by the formula:

[0182] $\mathrm{ATH}\left(F\right)=\frac{3.64}{F_{\mathrm{KHz}}^{0.8}}-6.5e^{-0.6{\left(F_{\mathrm{kHz}}-3.3\right)}^2}+0.001F_{\mathrm{kHz}}^4-3.37$

[0183] In some embodiments, as shown in FIGS. 9, 10, 12, and 14, a computational model can be used in place of explicit computation of the Earspring and Conductor equations, where the computational model is

[0184] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1}$
where A, B, C, D and E are parameters which have been fit against solutions to the combined Earspring and Conductor Equations, for each value of P₀ in 5 dB increments, and stored in a lookup table, such that for a given P₀, parameters for A, B, C, D, and E are extracted from the table, and then the computational model is used to compute ΔP given P.

[0185] In some embodiments it is also possible to use separate computational models for S(P) and for S⁻¹(S), with separate sets of parameters fit against the solutions for various boundary parameters and values of P₀.

[0186] In some embodiments as in FIGS. 9 and 10, the processing chain, from output of channels at the digital filter bank to adding gains to channels via the gain component, is a process that occurs in parallel for all channels of the audio signal, through the use of threaded processes, or through the use of a parallel computing device, such as a multi-core processing unit.

[0187] As shown in FIG. 15, in some embodiments the present invention may be implemented on a computer-readable medium 1070 storing a set of instructions executable by one or more processors, where the computer-readable medium 1070 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of channels which constitute a first audio signal, wherein the gains are applied using a modified form of parallel compression, where the compression parameters are dynamically varied in accordance with the noise, and parameters are selected to approximate the compression curves demanded by a psychoacoustic model, wherein said source signal is a channel of the audio signal corresponding to a particular frequency,

• • [0188] the computer-readable medium comprising:
    • [0189] a) code for solving for the power P(f) in Phons of the source signal of the source signal, at a frequency:
      • [0190] i. code for extracting a power value of the source signal at a frequency, in dBFS with a first power estimator block 1008, comprising:
        • [0191] 1. extracting the first audio signal x(t) 1002, said first audio signal comprising a plurality of channels corresponding to varying frequencies f, where said first audio signal is then transmitted to an Analog to Digital Converter 1004 thereby converting the signal into digital format;
        • [0192] 2. feeding the first audio signal through a First Frequency Analysis Module 1006 where the first audio signal is broken down into separate frequency components;
        • [0193] 3. identifying a particular source frequency component of the first audio signal corresponding to a frequency 1050;
        • [0194] 4. generating a power value for the frequency component, whereby said power value is then fed into a dB Converter to generate a power value P_dBFS(f) for the source in dBFS 1052;
      • [0195] ii. code for calculating a calibration offset to convert the power value P_dBFS(f) of the source signal in dBFS to a power value P_dBSPL(f) in dBSPL 1010 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12;
      • [0196] iii. code for converting the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation 1012, producing P(f);
    • [0197] b) code for solving for P₀(f) in Phons of the background signal, at a frequency:
      • [0198] i. code for extracting a power value of the background signal at the frequency, in dBFS with a second Power Estimator Block 1009, comprising:
        • [0199] 1. extracting a second audio signal x₀(t) 1003, said second audio signal comprising a plurality of channels corresponding to varying frequencies, where said second audio signal is then transmitted to a second Analog to Digital Converter 1005 thereby converting the signal into digital format;
        • [0200] 2. feeding the second audio signal through a Second Frequency Analysis Module 1007 where the second audio signal is broken down into separate components;
        • [0201] 3. identifying a particular background signal component, corresponding to a particular frequency 1050;
        • [0202] 4. obtaining a power value for the background signal component, whereby said power value is then fed into a dB Converter to generate a power value P_0—dBFS(f) for the background signal in dBFS 1052;
      • [0203] ii. code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dbSPL(f) in dBSPL using the calibration equation 1011 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12;
      • [0204] iii. code for converting the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation 1013, producing P₀(f);
    • [0205] c) code for determining parameters for the parallel compression module, including compression ratio, makeup gain, and threshold, based on the power value of the background signal,
    • [0206] d) code for applying parallel compression to the source signal, using the parameters previously determined,
    • [0207] e) code for reconstituting the first audio signal with the corrective gain to the source signal at the frequency, comprising combining, using a Frequency Synthesis Module 1018, the channels of the corrected signal, reconstituting the corrected signal, producing x_out(t);
    • [0208] f) code for feeding the corrected audio signal through a Digital to Analog Converter 1019;

[0209] In some embodiments, discrete sets of parameters for the parallel compression are found by dividing the range of possible noise volumes into discrete levels, where sets of parameters are found for each discrete noise loudness level, where the parameter estimation module 422 is a lookup table, where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.

[0210] In some embodiments, the parameter estimation module 422 is a polynomial continuous curve defining the relationship between the ambient noise loudness and each of the parameters, where polynomial regression is used to find parameters which define this curve, where these parameters are then stored in memory, so that during processing the parameters can then be computed using a polynomial equation.

[0211] In some embodiments, one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.

[0212] In another possible embodiment, the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
G=g0+(dP−T′)/R−dP

• • [0213] where dP is the difference between the signal loudness and noise loudness in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.

[0214] In some embodiments, a fixed set of parameters is used. The fixed parameters may fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.

[0215] In some embodiments, the First and Second Frequency Analysis Modules 406, 407 are composed of digital filter banks which are filters that break the signal down into various sub-band channels x(t) and x₀ (t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module 418 is a summer operation that combines the various channels into output xout(t).

[0216] In some embodiments, the First and second Frequency Analysis Modules 406,407 are digital filter banks composed of wavelet filters.

[0217] In some embodiments, the various sub-band channels correspond to critical bands of hearing.

[0218] In some embodiments, the First and Second Frequency Analysis Modules 406,407 are composed of FFT coprocessors that perform Fourier transforms on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module 418 is an FFT coprocessor that performs an inverse Fourier transform, resulting in output signal x_out(t).

[0219] In some embodiments, the frequency components may be grouped together to approximate sub-bands correspond to critical bands of hearing.

[0220] In some embodiments, the present invention features systems for audio correction. The systems may include but are not limited to mobile phones (and the like) and music players (and the like).

[0221] In some embodiments, the present invention may be multi-channel, such as in a stereo or surround sound system. In this case, gains are computed independently for each channel independently. These gains may be based on one central background noise estimate, or multiple noise estimates at each speaker. Each channel is then amplified by its corresponding gains so that the overall surround effect compensates for the presence of background noise.

[0222] In some embodiments, the present invention features an apparatus for correcting an audio signal. The apparatus, or audio source as claimed, (e.g., a mobile phone) may, for example, comprise a standard mobile phone receiver, a standard mobile phone microphone, and a standard mobile phone speaker, all of which are well known to one of ordinary skill in the art. The receiver can function to extract an amplitude of a source signal (e.g., an incoming conversation on a mobile phone) at given frequency (or within a frequency range) and the microphone can function to detect an amplitude of a background signal at a corresponding frequency (or within a frequency range).

[0223] In some embodiments, the systems of the present invention can evaluate sounds within pre-determined ranges of frequencies, e.g., any appropriate set or group of ranges. Microphones, and/or receivers and/or the like can collect information for the particular frequency range (the pre-determined frequency range). In some embodiments, a first range is 500 Hz and below, a second range is between 500 Hz and 2 kHz, and a third range is above 2 kHz. In some embodiments a first range is 1 kHz and below and a second range is above 1 kHz. The present invention is not limited to the aforementioned ranges.

[0224] In some embodiments, some elements of the above process flow might be performed using analog circuitry before being input to the microprocessor.

[0225] In some embodiments, Analog-to-digital and digital-to-analog converters are assumed to exist at interface points between analog elements and digital elements of any embodiment.

[0226] In some embodiments, programmable logic devices, including but not limited to FPGAs, may be used to implement parts of the processing shown in FIG. 3, with appropriate interfaces implemented between the microprocessor and the programmable logic devices.

[0227] In some embodiments, referring to the means of extracting amplitudes of source signals or background signals: at the output of each filter which selects a band of adjacent frequencies, an electrical circuit to effect extraction of amplitudes could be the use of a diode, or rectifier bridge, fed by each signal which, in turn, feeds a simple series resistor and a capacitor in shunt to create an integrator of the absolute value of the signal. The integrator would accumulate rectified current for some period of time before being applied to the loudness response curve selection circuit, after which the capacitor would be shorted to discharge the accumulated current, in anticipation of the next cycle of integration. Alternatively, the capacitor could be allowed to discharge continuously through a higher valued resistor in shunt, so that charge can be bled off more slowly than it is accumulated, thereby forming a running exponential average integrator.

[0228] In some embodiments, a more sophisticated version of amplitude estimation could send the signals through a squarer circuit ahead of the RC integrator in order to accumulate an estimate of the power in the signal. The resulting integrated current would be an estimate of the RMS power at the end of the integration period. This power estimate could then feed the gain correction curve selection circuit.

[0229] In some embodiments, the output of each integrator can be fed to an analog differencing amplifier (operational amplifier, or OpAmp) to subtract the estimated amplitude of the noise signal from the estimated amplitude of the signal of interest. The output of this difference amplifier can be used to establish a dynamic compressor threshold. The output of the compressor circuit (a voltage controlled amplifier, or VCA) can be added to the dry signal to effect a nonlinear correction. A simple linear compressor (VCA) adjusted with suitable make-up gain and compression ratio can produce, in sum with the dry signal, an approximation of the required nonlinear correction gain.

[0230] The present invention is not limited to the aforementioned examples of means of extracting amplitudes.

[0231] Without wishing to limit the present invention to any theory or mechanism, it is believed that the present invention is advantageous because the present invention features a maximum output volume, for example the source signal will be amplified to only a certain degree. This can help protect against damage to the user's hearing. Furthermore, the systems frequently detect the background noise (e.g., 10-40 times per second, 1,000 times per second, etc.), allowing for the systems to react quickly to changes in the background noise. Furthermore, the systems utilize non-linear corrections, thus, loud signals are not amplified much if at all.

[0232] In some embodiments, the systems of the present invention can allow for microphone sensitivity to be altered. For example, a particular user may wish to have more correction or less correction. For example, the microphone may be in someone's pocket, or else exposed directly to the ambient noise sounds. If the microphone becomes shielded, it will produce a weaker response to the noise than if it were directly exposed. Hence the microphone gain can be increased to compensate for the shielding. Also, each of us has a slightly different response to noise levels, some people being more sensitive to it than others. A variable gain on the microphone can help to compensate for individual differences of loudness perception. Finally microphones exhibit varying characteristics depending on their specific construction. A variable microphone gain in the software can help to normalize the response from different microphones.

[0233] In some embodiments, the systems of the present invention can allow for signals to be corrected prior to transmission or recording, such that ambient noise in the recorded or received signal is retained, but the spectral composition of the primary source signal is modified to compensate, as an alternative to standard noise reduction techniques.

[0234] In some embodiments, steps for converting analog audio input signals to digital input signals can be bypassed where the invention utilizes digital input audio sources capable of receiving digital audio signals and transmitting the same to the processor.

[0235] As used herein, the term “hearing impaired” refers to physical hearing impairment, as in the traditional sense of hearing loss, and/or, can also refer to temporary hearing loss caused by a background noise or other temporary hearing loss factors. One of the novel discoveries addressed by the present invention is that a individual with normal hearing can in fact experience hearing impairment due to background noise, or background signals interfering with source noise or source signals.

[0236] In some embodiments, the present invention comprises an apparatus 100, effective in enhancing a first audio signal in the face of ambient background noise, by applying corrective gains to a plurality of frequency components which constitute the first audio signal, wherein the value of the gains are computed using a nonlinear psychoacoustic model of loudness as a function of sound intensity, as subjectively experienced by a normal human, where this psychoacoustic model takes as inputs the sound intensity of the audio signal P as well as the ambient background noise P₀, wherein corrective gains are generated for each frequency component by solving for a correction gain ΔP for the first audio signal which will suffice to raise the volume of each frequency component of the first audio signal so that the user will perceive it above background noise, by the loudness amount equivalent to how the first audio signal would be perceived in a noiseless environment.

[0237] In some embodiments, the apparatus comprises a first audio source 02, configured to extract a first audio signal where the first audio signal is a broadband audio signal of interest, the first audio source operatively connected to a microprocessor 01.

[0238] In some embodiments, the apparatus comprises a second audio source 03, configured for extracting a second audio signal comprising ambient broadband audio in the environment, the second audio source operatively connected to the microprocessor 01.

[0239] In some embodiments, the apparatus comprises the microprocessor 01, configured to execute applications for:

• • [0240] i. a first Analog to Digital converter 04;
  • [0241] ii. a second Analog to Digital converter 05;
  • [0242] iii. a first Frequency Analysis Module 06;
  • [0243] iv. a second Frequency Analysis Module 07;
  • [0244] v. a first Power Estimator Block 08;
  • [0245] vi. a second Power Estimator Block 09;
  • [0246] vii. a first dBFS to dBSPL Converter 10;
  • [0247] viii. a second dBFS to dBSPL Converter 11;
  • [0248] ix. a first dBSPL to dB Phon Converter 12;
  • [0249] x. a second dBSPL to dB Phon Converter 13;
  • [0250] xi. a Psychoacoustic model 22;
  • [0251] xii. a Gain Block component 17;
  • [0252] xiii. a Frequency Synthesis Module 18; and
  • [0253] xiv. a Digital to Analog Converter 19.

[0254] In some embodiments, the apparatus operates according to the following steps:

[0255] (i) the first audio source 02 extracts the first audio signal x(t), where the first audio signal is then transmitted to the first Analog to Digital Converter 04 thereby converting the first audio signal into digital format; and, (ii) the second audio source 03 extracts a second audio signal x₀(t), where said second audio signal is then transmitted to the second Analog to Digital Converter 05 thereby converting the second audio signal into digital format,
whereupon the first audio signal and the second audio signal are fed, respectively, through the First Frequency Analysis Module 06 and Second Frequency Analysis Module 07 breaking down each respective audio signal into separate pluralities of frequency components,
whereupon, for each frequency component of the first audio signal, the apparatus 100 identifies a source signal and a background signal, where said source signal is a frequency component of the first audio signal and the background signal is the frequency component of the second audio signal 50,
whereupon both the source signal and the background signal are fed, respectively, through the first Power Estimator 08 and the Second Power Estimator 09, which compute power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS 52,
whereupon (i) the microprocessor 01 completes a calibration offset such that a first dBFS to dBSPL Converter 10 converts the power value of the source signal from dBFS to dBSPL, and (ii) the microprocessor 01 completes a calibration offset such that a second dBFS to dBSPL Converter 11 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) for the source signal and P_0—dBSPL(f) for the background signal,
whereupon (i) a first dBSPL-dB Phon Converter 12 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 13 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f),
whereupon the microprocessor 01 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the psychoacoustic model 22 whereupon the microprocessor 01 calculates ΔP(f) in dBPhons,
whereupon the microprocessor 01 calculates a correction gain of power to the source signal G(f), at the frequency, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 17 which utilizes the Gain equation,
whereupon the apparatus 100 reconstitutes the first audio signal with the corrective gains to the source signals corresponding to each of the frequency components, comprising:

• • [0256] (i) multiplying the source signal corresponding to each of the frequency components according to the corresponding corrective gain ΔP(f), producing a corrected source signal,
  • [0257] (ii) combining the corrected source signals using the Frequency Synthesis Module 18, to producing a corrected first audio signal
    whereupon the corrected first audio signal is fed through the Digital to Analog Converter 19, resulting in analog signal x_out(t).

[0258] In some embodiments, the apparatus comprises an apparatus 200, effective in enhancing an audio signal, by applying corrective gains to a plurality of frequency components which constitute a first audio signal, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to a signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
G=g0+(P−T)*(1/R−1),
where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behaviour, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module.

[0259] In some embodiments, the apparatus comprises a first audio source 202, configured to extract a first audio signal, where the first audio signal is a broadband audio signal of interest, the first audio source operatively connected to a microprocessor 201.

[0260] In some embodiments, the apparatus comprises a second audio source 203, configured for extracting a second audio signal from the environment where the second audio signal is ambient noise, the second audio source operatively connected to the microprocessor 201.

[0261] In some embodiments, the apparatus comprises the microprocessor 201, operatively connected to a Speaker 220, and an Amplifier 219; the microprocessor 201 configured to execute applications for:

• • [0262] i. a first Analog to Digital converter 204,
  • [0263] ii. a second Analog to Digital converter 205,
  • [0264] iii. a first frequency analysis module 206, configured to extract a plurality of frequency components from the first audio signal,
  • [0265] iv. a second frequency analysis module 207, configured to extract a plurality of frequency components from the second audio signal,
  • [0266] v. a first power estimation block 208, configured to calculate the sound intensity in decibels of each frequency component of the first audio signal 002,
  • [0267] vi. a second power estimation block 209, configured to calculate the sound intensity of decibels of each frequency component of the second audio signal 003,
  • [0268] vii. a first dBFS to dBSPL Converter 210,
  • [0269] viii. a second dBFS to dBSPL Converter 211,
  • [0270] ix. a parameter estimation module, which determines parameter settings for a parallel compression module, according to information stored in memory,
  • [0271] x. the parallel compression module 215, parameterized by threshold T, compression ratio R, and makeup gain g0, where the parameters may by dynamically varied, where the parallel compression module is comprised of:
    • [0272] a. a linear compression module 214, which applies gains to the first audio signal according to the equation
      G=g0+(P−T)*(1/R+1),
    • [0273] where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
    • [0274] b. a summer, which combines the resulting amplified first audio signal component with the original first audio signal component,
  • [0275] xi. a Frequency Synthesis Module 216; and
  • [0276] xii. a Digital to Analog Converter 219;

[0277] In some embodiments, the apparatus operates according to the following steps:

[0278] (i) the first audio source 202 extracts the first audio signal x(t), where the first audio signal is then transmitted to the first Analog to Digital Converter 204 thereby converting the first audio signal into digital format; and, (ii) the second audio source 203 extracts the second audio signal x₀(t), where said second audio signal is then transmitted to the second Analog to Digital Converter 205 thereby converting the second audio signal into digital format,
whereupon the first audio signal and the second audio signal are fed, respectively, through the First Frequency Analysis Module 206 and Second Frequency Analysis Module 207 breaking down each respective audio signal into separate components,
whereupon, for each frequency component, the apparatus 200 identifies a source signal and a background signal corresponding to a frequency component 250, and said signals are thereafter transmitted to the microprocessor 201,
whereupon, for each frequency component, both the source signal and the background signal are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the signals, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS 252,
whereupon, for each frequency component, and at each instant, the microprocessor, using the power estimates of the source signal and background signal, obtains parameters for the parallel compression, depending on the sound intensity of the background signal at each instant,
whereupon the microprocessor, applies the parameters to a parallel compression module,
whereupon the parallel compression module applies a linear compression, according to the selected parameters, to each of the frequency components of the first audio signal,
whereupon the parallel compression module sums each frequency component of the first audio signal with the corresponding compressed audio signal,
whereupon the apparatus 200 reconstitutes the plurality of frequency components of the corrected first audio signal using the Frequency Synthesis Module 218, to produce a corrected first audio signal,
whereupon the corrected first audio signal is fed through the Digital to Analog Converter 219, resulting in analog signal x_out(t).

[0279] In some embodiments, the First and Second Frequency Analysis Modules are composed of digital filter banks which are IIR filters that break the signal down into various sub-band channels x(t) and x₀(t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module is a summer operation that combines the various channels into output x_out(t).

[0280] In some embodiments, the various sub-band channels correspond to critical bands of hearing.

[0281] In some embodiments, the First and second Frequency Analysis Modules are digital filter banks composed of wavelet filters.

[0282] In some embodiments, the First and Second Frequency Analysis Modules are composed of FFT coprocessors that perform Fourier transforms on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module is an FFT coprocessor that performs an inverse Fourier transform, resulting in output signal x_out(t).

[0283] In some embodiments, the frequency components are grouped together to approximate sub-bands correspond to critical bands of hearing.

[0284] In some embodiments, the psychoacoustic model 22 consists of a forward model which computes subjectively perceived Sones in terms of Phons, and an inverse model which computes Phons in terms of Sones, allowing the Psychoacoustic model to be combined into the Conductor Equation, the Conductor Equation 16 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,

• • [0285] where this equation may be stated as the desired gain in Phons being the Phons amplification necessary to amplify the sound over background noise so that it's subjectively perceived loudness over background noise is the same as the subjectively perceived loudness without the noise, where the function S(P) is found by solving the Earspring equation under experimental boundary conditions, where the Earspring equation 14,15 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), which is a harmonic oscillator model of the signal processing performed by the ear, the cochlea, and immediate processing by the auditory cortex, where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration for a given frequency channel, identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor, where for a particular frequency of the driving force F(t), which represents a pure sinusoid of intensity P, a solution y(t) to this equation can be found, which can then be converted to <y²>, which represents the perceived loudness S in Sones, resulting in a function S(P),

[0286] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10},$

• • [0287] where S(P) is the perceived loudness in Sones which the listener experiences for a particular sub-band, and P is the sound intensity impinging on the ear in dBPhons for a particular subband, where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency of the earspring, where the constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions.

[0288] In some embodiments, the set of boundary parameters used to obtain solutions to the Earspring Equation comprises: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz. In some embodiments, the aforementioned set of experimental boundary conditions comprise: (i) a Sones ratio between the hearing threshold level and the reference level of ˜20-30 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 50-75 cents as tones range in intensity from ˜0 dBSPL to ˜40 dBSPL near 1 kHz. In some embodiments, the aforementioned set of experimental boundary conditions comprise: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40-60 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 50-100 cents as tones range in intensity from ˜0 dBSPL to ˜100 dBSPL near 1 kHz.

[0289] In some embodiments, a computational model is used to estimate Earspring equation solutions S(P) and S⁻¹(S) consisting of quadratic functions such as

[0290] $S\left(P\right)\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring equation and are stored in long term memory.

[0291] In some embodiments, the psychoacoustic model is replaced by a computational model 124 which is used to approximate the output of the combined psychoacoustic model and conductor equation, where the computation model comprises a quadratic function

[0292] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Conductor equation 116, for various values of the threshold elevation P₀ and are stored in a lookup table 123.

[0293] In some embodiments, the psychoacoustic model 122 is replaced by any numerical function approximator 124 that is fit against the Earspring 114, 115 and Conductor equations 116, and approximates their solutions, where these approximators may include, but are not limited to linear regression, neural networks, polynomial regression, logistic regression, kernel regression, factor analysis, spline fitting, piecewise curve fitting, radial basis functions, and any other suitable approximator now known or later discovered.

[0294] In some embodiments, discrete sets of parameters are found by dividing the range of possible noise volumes into discrete levels, and where sets of parameters are found for each discrete noise loudness level, where the parameter estimation module 111 is a lookup table, where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.

[0295] In some embodiments, the parameter estimation module 111 is a polynomial continuous curve defining the relationship between the ambient noise loudness and each of the parameters, where polynomial regression is used to find parameters which define this curve, where these parameters are then stored in memory, so that during processing the parameters can then be computed using a polynomial equation.

[0296] In some embodiments, one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.

[0297] In some embodiments, the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
G=g0+(dP−T′)/R−dP

• • [0298] where dP is the difference between the signal loudness and noise loudness in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.

[0299] In some embodiments, a fixed set of parameters is used.

[0300] In some embodiments, the fixed parameters fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset. In some embodiments, the fixed parameters fall in the range of 0 to 10 for compression ratio, 0 dB to 30 dB for makeup gain, and −25 to −40 dB for threshold offset. In some embodiments, the fixed parameters fall in the range of 6 to 50 for compression ratio, 10 dB to 50 dB for makeup gain, and −40 to 10 dB for threshold offset.

[0301] In some embodiments, the apparatus is a mobile phone.

[0302] In some embodiments, the present invention comprises a method of obtaining the solution to the Earspring equation 114,115 and it's inverse function in a computationally efficient manner, using a computational model in which various parameters of the model are fit against a variety of solutions of the Earspring equation, and it's inverse, for various input amplitudes and/or frequencies, where the parameters are stored within a long term memory component. In one exemplary embodiment, the computational model that estimates S(P) may consist of a quadratic function

[0303] $S\left(P\right)\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring equation and are stored in long term memory.

[0304] In some embodiments, the present invention comprises a method of obtaining corrective gains ΔP corresponding to a previously determined spectrum of threshold elevations, using the Conductor Equation 116 and perceived loudness models S(P) and S⁻¹(S) based on solutions to the Earspring equation, the Conductor Equation 116 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,

• • [0305] where S(P)=a Sones equation from an EarSpring equation 114,115 solution for the given frequency, as a function of Phons Pressure, and S⁻¹(S) is an inverse function of this solution that gives the sound intensity P in Phons needed for a normal listener to perceive the sound at loudness S in Sones, where P(f)=a Phons amplitude of the frequency component at f of a source signal of interest, and P₀(f)=a Phons amplitude of the frequency component at f of a background signal of interest, where Sones is a measure of perceived loudness under subjective experience and Phons is a measure of objective sound intensity, where the Earspring equation 114,115 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), which is a harmonic oscillator model of the signal processing performed by the ear, the cochlea, and immediate processing by the auditory cortex, where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration for a given frequency channel, identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor,
  • [0306] where for a particular frequency of the driving force F(t), which represents a pure sinusoid of intensity P, a solution y(t) to this equation can be found, which can then be converted to <y²>, which as the output of the equation represents the perceived loudness S in Sones, producing a function S(P),

[0307] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10},$

• • [0308] where S(P) is the perceived loudness in Sones which the listener experiences for a particular frequency, and P is the sound intensity impinging on the ear in dBPhons for a particular frequency, where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency of the earspring, where the constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions.

[0309] In some embodiments, the set of boundary parameters comprises: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz.

[0310] In some embodiments, the gains ΔP are obtained in a computationally efficient manner, using a single computational model 124 that combines S(P) and its inverse in the Conductor Equation 116 to obtain a single computational model

[0311] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the Conductor equation and are stored in a lookup table 123.

[0312] In some embodiments, the psychoacoustic model is a computational model 124 which may be any numerical function approximator that is fit against the Earspring and Conductor equation and approximates their solutions, where these approximators may include, but are not limited to linear regression, neural networks, polynomial regression, logistic regression, kernel regression, factor analysis, spline fitting, piecewise curve fitting, radial basis functions, and any other suitable approximator now known or later discovered.

[0313] In some embodiments, the present invention comprises a method of using parallel compression to approximate a desired non-linear dynamic range compression curve, wherein a compression curve defines the output loudness in decibels versus the input loudness in decibels of a signal of interest, wherein the compression is implemented by applying a gain G to the signal of interest according to the following equation:
G=g0+(P−T)*(1/R−1),
where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
whereupon the linearly compressed signal is added together with the original signal, producing a parallel compressed output signal, where the overall amplitude of the input versus the output signal defines a parallel compression curve, where g0, T, and R are thus parameters that vary the shape of the parallel compression curve, and P is the input loudness of the signal of interest,
where the method is comprised of the following steps:

• • [0314] a) obtaining parameters that shape the parallel compression curve, by fitting compression curves using these parameters against a plurality of desired compression curves and selecting the parameters that minimize the difference between the compression curve of the parallel compression module and each of the desired compression curves,
  • [0315] b) storing the optimal sets of parameters in a memory device,
  • [0316] c) during real-time processing, at each time step:
    • [0317] i. retrieving a set of parameters, depending on the desired compression curve to be used,
    • [0318] ii. applying the set of parameters, to the parallel compression module; and processing a first audio signal using the time-varying compression parameters, in the parallel compression module, where the parallel compression module is performed by:
      • [0319] 1. applying linear compression to the first audio signal according to the equation G=g0+(P−T)*(1/R+1), producing a linearly compressed
      • [0320] 2. summing the linearly compressed audio signal with the first audio signal

[0321] In some embodiments, the plurality of desired dynamic range compression curves approximate the ideal compression curves needed to compensate for a plurality of ambient noise signal intensities, so that the derived parallel compression parameters correspond to a plurality of ambient noise loudness levels.

[0322] In some embodiments, discrete sets of parameters are found by dividing the range of possible noise volumes into discrete levels, and where sets of parameters are found for each discrete noise loudness level, where the sets of parallel compression parameters are stored in a lookup table, and where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.

[0323] In some embodiments, polynomial regression is used to fit a continuous curve between the ambient noise loudness and each of the parameters, where the parameters of this curve are stored, and where during real time processing the ambient noise loudness is measured and applied to the polynomial function to obtain each parameter T, R, and g0.

[0324] In some embodiments, one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.

[0325] In some embodiments, the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
G=g0+(dP−T′)/R−dP
where dP is the difference between the signal loudness and noise loudness in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.

[0326] In some embodiments, a fixed set of parameters is used.

[0327] In some embodiments, the fixed parameters fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.

[0328] In some embodiments, the signal is divided into a plurality of sub-bands and parallel compression is applied independently to each channel.

[0329] In some embodiments, a threshold setting for the parallel compression is set according to a current threshold elevation value, wherein the threshold elevation is a function of a second audio signal.

[0330] In some embodiments, the present invention comprises a computer-readable medium 670 storing a set of instructions executable by one or more processors, where the computer-readable medium 670 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of frequency components which constitute a first audio signal, wherein corrective gains are generated for each frequency component by solving for a correction value of ΔP for a source signal, wherein said source signal is a frequency component of the first audio signal corresponding to a particular frequency component, where ΔP is derived from solving for the output of a psychoacoustic model 622.

[0331] In some embodiments, the computer-readable medium comprises:

• • [0332] a) code for solving for the power P(f) in Phons of the source signal of the first audio signal, at each of the frequency components of the first audio signal:
    • [0333] i. code for extracting a power value of the source signal at a frequency component, in dBFS with a first power estimator block 608, comprising:
      • [0334] (A) extracting the first audio signal x(t) 602, said first audio signal comprising a plurality of frequency components, where said first audio signal is then transmitted to an Analog to Digital Converter 604 thereby converting the signal into digital format,
      • [0335] (B) feeding the first audio signal through a First Frequency Analysis Module 606 where the first audio signal is broken down into separate frequency components,
      • [0336] (C) identifying a the source signal of the first audio signal corresponding to a frequency component 650,
      • [0337] (D) generating a power value for the source signal, whereby said power value is then fed into a dB Converter to generate a power value P_dBFS(f) for the source signal in dBFS 652,
    • [0338] ii. code for calculating a calibration offset to convert the power value P_dBFS(f) of the source signal in dBFS to a power value P_dBSPL(f) in dBSPL 610 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,
    • [0339] iii. code for converting the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation 612, producing P(f),
  • [0340] b) code for solving for P₀(f) in Phons of the background signal, at the frequency component:
    • [0341] i. code for extracting a power value of the background signal at the frequency component, in dBFS with a second Power Estimator Block 609, comprising:
      • [0342] (A) extracting a second audio signal x₀(t) 603, said second audio signal comprising a plurality of frequency components, where said second audio signal is then transmitted to a second Analog to Digital Converter 605 thereby converting the signal into digital format,
      • [0343] (B) feeding the second audio signal through a Second Frequency Analysis Module 607 where the second audio signal is broken down into separate frequency components,
      • [0344] (C) identifying a particular background signal, corresponding to the frequency component 650,
      • [0345] (D) obtaining a power value for the background signal, whereby said power value is then fed into a dB Converter to generate a power value P_0—dBFS(f) for the background signal in dBFS 652,
    • [0346] ii. code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dbSPL(f) in dBSPL using the calibration equation 611 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,
    • [0347] iii. code for converting the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation 613, producing P₀(f),
  • [0348] c) code for using the psychoacoustic model 622 to solve for ΔP(f) using P(f) and P₀(f);
  • [0349] d) code for calculating a correction gain of power G(f), for a frequency f, in dBSPL based on the value of ΔP(f) in dB Phons, using a Gain equation 617;
  • [0350] e) code for reconstituting the first audio signal with corrective gains to the source signals at each of the frequency components of the first audio signal, comprising
    • [0351] i. feeding the correction gain to an amplifier wherein the amplifier amplifies the source signal at the frequency component by the corrective gain G(f) 653,
    • [0352] ii. combining, using a Frequency Synthesis Module 618, the source signals of the corrected signal, producing a corrected first audio signal, x_out(t),
  • [0353] f) code for feeding the corrected first audio signal through a Digital to Analog Converter 619.

[0354] In some embodiments, the computer-readable medium comprises a computer-readable medium 70 storing a set of instructions executable by one or more processors, where the computer-readable medium 1070 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of frequency components which constitute a first audio signal, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the compression is implemented by applying a gain G to the signal of interest according to the following equation:
G=g0+(P−T)*(1/R−1)

• • [0355] where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
    • [0356] whereupon the linearly compressed signal is added to the first audio signal, where the overall amplitude of the input versus the output signal defines a parallel compression curve, where g0, T, and R are thus parameters that vary the shape of the resulting compression curve, where the psychoacoustic model takes as inputs a source signal loudness and a noise signal loudness, where the model computes the gain in sound loudness at a given frequency component required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness level, are determined beforehand and stored in memory, where a different set of parameters is retrieved at each instant depending on the ambient noise signal intensity at each frequency component, where the dynamically changing noise signal intensity results in a dynamically changing selection of parameters for the parallel compression module.

[0357] In some embodiments, the computer-readable medium comprises:

• • [0358] a) code for, at each frequency component, extracting a power value of the source signal at a frequency component, in dBFS with a first power estimator block 1008; comprising:
    • [0359] i. extracting the first audio signal x(t) 1002, said first audio signal comprising a plurality of frequency components, where said first audio signal is then transmitted to an Analog to Digital Converter 1004 thereby converting the signal into digital format,
    • [0360] ii. feeding the first audio signal through a First Frequency Analysis Module 1006 where the first audio signal is broken down into separate frequency components,
    • [0361] iii. identifying a particular source signal of the first audio signal corresponding to a frequency component 1050,
    • [0362] iv. generating a power value for the source signal 1051, whereby said power value is then fed into a dB Converter to generate a power value P_dBFS(f) for the source signal in dBFS 1052,
    • [0363] v. code for calculating a calibration offset to convert the power value P_dBFS(f) of the source signal in dBFS to a power value P_dBSPL(f) in dBSPL 1010 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,
  • [0364] b) code for extracting a power value of the background signal at the frequency component, in dBFS with a second Power Estimator Block 1009; comprising:
    • [0365] i. extracting a second audio signal x₀(t) 1002, said second audio signal comprising a plurality of frequency components, where said second audio signal is then transmitted to a second Analog to Digital Converter 1004 thereby converting the signal into digital format,
    • [0366] ii. feeding the second audio signal through a Second Frequency Analysis Module 1006 where the second audio signal is broken down into separate frequency components,
    • [0367] iii. identifying a particular background signal, corresponding to the frequency component 1050,
    • [0368] iv. obtaining a power value for the background signal 1051, whereby said power value is then fed into a dB Converter to generate a power value P_0—dBFS(f) for the background signal in dBFS 1052,
    • [0369] v. code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dbSPL(f) in dBSPL using the calibration equation 1010 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,
  • [0370] c) code for determining parameters for the parallel compression module 1022, including compression ratio, makeup gain, and threshold, based on the power value of the background signal, where the parameters have previously been fit against solutions to the Earspring and conductor equations, for a variety of background signal power values, and where the parameters may be retrieved at runtime, given a background signal power;
  • [0371] d) code for applying parallel compression to the source signal 1023, using the stored parameters now retrieved, thereby amplifying the source signal according to the desired compression;
  • [0372] e) code for reconstituting the audio signal with the corrective gain to the source signal at the frequency component, comprising combining, using a Frequency Synthesis Module 1018, the parallel compressed frequency components of the audio signal, producing a corrected audio signal, xout(t);

[0373] In some embodiments, the First and Second Frequency Analysis Modules are composed of Digital Filter Banks which are filters that break the signal down into varios channels x(t) and x₀ (t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module is a summer operation that combines the various channels into output x_out(t).

[0374] In some embodiments, the First and second Frequency Analysis Modules are digital filter banks composed of wavelet filters.

[0375] In some embodiments, the various channels correspond to critical bands of hearing.

[0376] In some embodiments, the First and Second Frequency Analysis Modules perform a Fourier transform on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module performs an inverse Fourier transform, resulting in output signal x_out(t).

[0377] In some embodiments, the psychoacoustic model 122 consists of a forward model which computes subjectively perceived Sones in terms of Phons, and an inverse model which computes Phons in terms of Sones, allowing the Psychoacoustic model to be combined into the Conductor Equation, the Conductor Equation 116 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,

• • [0378] where this equation may be stated as the desired gain in Phons being the Phons amplification necessary to amplify the sound over background noise so that it's subjectively perceived loudness over background noise is the same as the subjectively perceived loudness without the noise, minus some correcting factor, where the function S(P) is found by solving the Earspring equation under experimental boundary conditions, where the Earspring equation 114,115 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), which is a harmonic oscillator model of the signal processing performed by the ear, the cochlea, and immediate processing by the auditory cortex, where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration for a given frequency channel, identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor, where for a particular frequency of the driving force F(t), which represents a pure sinusoid of intensity P, a solution y(t) to this equation can be found, which can then be converted to <y²>, which as the output of the equation represents the perceived loudness S in Sones, producing a function S(P),

[0379] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10},$

• • [0380] where S(P) is the perceived loudness in Sones which the listener experiences for a particular frequency, and P is the sound intensity impinging on the ear in dBPhons for a particular frequency, where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency of the earspring, where the constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions.

[0381] In some embodiments, the set of boundary parameters used to obtain solutions to the Earspring Equation comprises: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz.

[0382] In some embodiments, the computational model 122 is used to estimate S(P) and S⁻¹(S) consisting of a quadratic function

[0383] $S\left(P\right)\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring and Conductor equations and are stored in long term memory.

[0384] In some embodiments, the computational model 122 is used to estimate ΔP consists of a quadratic function

[0385] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring and Conductor equations and are stored in a lookup table 123.

[0386] In some embodiments, the computational model 122 is any numerical function approximator that is fit against the Earspring and Conductor equation and approximates their solutions, where these approximators may include, but are not limited to linear regression, neural networks, polynomial regression, logistic regression, kernel regression, factor analysis, spline fitting, piecewise curve fitting, radial basis functions, and any other suitable approximator now known or later discovered.

[0387] In some embodiments, the first audio signal comprises a plurality of frequency components and wherein a corrective gain corresponding to each respective component is processed and applied to each component in parallel, before summation.

[0388] In some embodiments, the first audio signal contains more than one frequency component and steps (a) through (d) are repeated in series for all frequencies in the spectrum, before proceeding to step (e).

[0389] In some embodiments, the first audio signal contains more than one frequency component and steps (a) through (d) are in part computed in parallel and in part computed in series, before proceeding to step (e).

[0390] In some embodiments, either the Earspring equation or its inverse are estimated using separate computational models as in claim 5, replacing either step (a)(vi) or (a)(vii), respectively.

[0391] In some embodiments, the dBSPL to dBPhon Conversion equation 12,13 is:

[0392] $P_{\mathrm{dBPhon}}\left(F\right)=\frac{P_{\mathrm{dBSPL}}\left(F\right)-\mathrm{ATH}\left(F\right)}{1+\mathrm{kATH}\left(F\right)}$

• • [0393] where
  • [0394] F=frequency
  • [0395] ATH(F)=Absolute Threshold of Hearing
  • [0396] K=− 1/120 when ATH(F)>0
    • [0397] =+ 1/240 when ATH(F)<0

[0398] In some embodiments the Gain equation 17 is:
ΔP_dBSPL(F)=ΔP_dBPhon(F)(1+kATH(F))

• • [0399] where
  • [0400] F=frequency
  • [0401] ATH(F)=Absolute Threshold of Hearing
  • [0402] K=− 1/120 when ATH(F)>0
    • [0403] =+ 1/240 when ATH(F)<0

[0404] In some embodiments, the absolute threshold of hearing is computed by the formula:

[0405] $\mathrm{ATH}\left(F\right)=\frac{3.64}{F_{\mathrm{KHz}}^{0.8}}-6.5e^{-0.6{\left(F_{\mathrm{kHz}}-3.3\right)}^2}+0.001F_{\mathrm{kHz}}^4-3.37$

[0406] In some embodiments, discrete sets of parameters are found by dividing the range of possible noise signal intensities into discrete levels, and where sets of parameters are found for each discrete noise intensity level, where the parameter estimation module 111 is a lookup table, where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.

[0407] In some embodiments, the parameter estimation module 111 is a polynomial continuous curve defining the relationship between the ambient noise loudness and each of the parameters, where polynomial regression is used to find parameters which define this curve, where these parameters are then stored in memory.

[0408] In some embodiments, one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.

[0409] In some embodiments, the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
G=g0+(dP−T′)/R−dP
where dP is the difference between the signal intensity and noise intensity in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.

[0410] In some embodiments, a fixed set of parameters is used.

[0411] In some embodiments, the fixed parameters fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.

[0412] In some embodiments, there are a plurality of source audio signals corresponding to the various signals of a stereo or surround sound audio system, and where gains for each channel are independently computed and applied to each channel.

[0413] In some embodiments, there are a plurality of source audio signals corresponding to the various signals of a stereo or surround sound audio system, and a plurality of background noise signals, one for each channel, and where gains for each channel are independently computed, according to the source and noise signal for each channel, and applied to each channel.

[0414] In some embodiments, the microprocessor is a DSP chip.

[0415] As used herein, the term “about” refers to plus or minus 10% of the referenced number. For example, an embodiment wherein the frequency is about 1,000 Hz includes an embodiment wherein the frequency is 900 to 1,100 Hz.

[0416] Various modifications of the invention, in addition to those described herein, will be apparent to those skilled in the art from the foregoing description. Such modifications are also intended to fall within the scope of the invention. Each reference cited in the present application is incorporated herein by reference in its entirety.

[0417] Although there has been shown and described the preferred embodiment of the present invention, it will be readily apparent to those skilled in the art that modifications may be made thereto which do not exceed the scope of the invention.

These are not Claims, they are Additional Disclosure:

[0418] 1. An apparatus 100, effective in enhancing a first audio signal in the face of ambient background noise, by applying corrective gains to a plurality of frequency components which constitute the first audio signal, wherein the value of the gains are computed using a nonlinear psychoacoustic model of loudness as a function of sound intensity, as subjectively experienced by a normal human, where this psychoacoustic model takes as inputs the sound intensity of the audio signal P as well as the ambient background noise P₀, wherein corrective gains are generated for each frequency component by solving for a correction gain ΔP for the first audio signal which will suffice to raise the volume of each frequency component of the first audio signal so that the user will perceive it above background noise, by the loudness amount equivalent to how the first audio signal would be perceived in a noiseless environment,
the apparatus 100 comprising:

[0419] a. a first audio source 02, configured to extract a first audio signal where the first audio signal is a broadband audio signal of interest, the first audio source operatively connected to a microprocessor 01;

[0420] b. a second audio source 03, configured for extracting a second audio signal comprising ambient broadband audio in the environment, the second audio source operatively connected to the microprocessor 01; the microprocessor 01, configured to execute applications for:

• • [0421] i. a first Analog to Digital converter 04;
  • [0422] ii. a second Analog to Digital converter 05;
  • [0423] iii. a first Frequency Analysis Module 06;
  • [0424] iv. a second Frequency Analysis Module 07;
  • [0425] v. a first Power Estimator Block 08;
  • [0426] vi. a second Power Estimator Block 09;
  • [0427] vii. a first dBFS to dBSPL Converter 10;
  • [0428] viii. a second dBFS to dBSPL Converter 11;
  • [0429] ix. a first dBSPL to dB Phon Converter 12;
  • [0430] x. a second dBSPL to dB Phon Converter 13;
  • [0431] xi. a Psychoacoustic model 22;
  • [0432] xii. a Gain Block component 17;
  • [0433] xiii. a Frequency Synthesis Module 18; and
  • [0434] xiv. a Digital to Analog Converter 19;
    wherein (i) the first audio source 02 extracts the first audio signal x(t), where the first audio signal is then transmitted to the first Analog to Digital Converter 04 thereby converting the first audio signal into digital format; and, (ii) the second audio source 03 extracts a second audio signal x₀(t), where said second audio signal is then transmitted to the second Analog to Digital Converter 05 thereby converting the second audio signal into digital format,
    whereupon the first audio signal and the second audio signal are fed, respectively, through the First Frequency Analysis Module 06 and Second Frequency Analysis Module 07 breaking down each respective audio signal into separate pluralities of frequency components,
    whereupon, for each frequency component of the first audio signal, the apparatus 100 identifies a source signal and a background signal, where said source signal is a frequency component of the first audio signal and the background signal is the frequency component of the second audio signal 50,
    whereupon both the source signal and the background signal are fed, respectively, through the first Power Estimator 08 and the Second Power Estimator 09, which compute power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS 52,
    whereupon (i) the microprocessor 01 completes a calibration offset such that a first dBFS to dBSPL Converter 10 converts the power value of the source signal from dBFS to dBSPL, and (ii) the microprocessor 01 completes a calibration offset such that a second dBFS to dBSPL Converter 11 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) for the source signal and P_0—dBSPL(f) for the background signal,
    whereupon (i) a first dBSPL-dB Phon Converter 12 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 13 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f),
    whereupon the microprocessor 01 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the psychoacoustic model 22 whereupon the microprocessor 01 calculates ΔP(f) in dBPhons,
    whereupon the microprocessor 01 calculates a correction gain of power to the source signal G(f), at the frequency, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 17 which utilizes the Gain equation,
    whereupon the apparatus 100 reconstitutes the first audio signal with the corrective gains to the source signals corresponding to each of the frequency components, comprising:
  • [0435] (i) multiplying the source signal corresponding to each of the frequency components according to the corresponding corrective gain ΔP(f), producing a corrected source signal,
  • [0436] (ii) combining the corrected source signals using the Frequency Synthesis Module 18, to producing a corrected first audio signal
    whereupon the corrected first audio signal is fed through the Digital to Analog Converter 19, resulting in analog signal x_out(t).
    2. An apparatus 200, effective in enhancing an audio signal, by applying corrective gains to a plurality of frequency components which constitute a first audio signal, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to a signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
    G=g0+(P−T)*(1/R−1),
    where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behavior, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module, the apparatus 200 comprising:

[0437] a) a first audio source 202, configured to extract a first audio signal, where the first audio signal is a broadband audio signal of interest, the first audio source operatively connected to a microprocessor 201;

[0438] b) a second audio source 203, configured for extracting a second audio signal from the environment where the second audio signal is ambient noise, the second audio source operatively connected to the microprocessor 201;

[0439] c) the microprocessor 201, operatively connected to a Speaker 220, and an Amplifier 219; the microprocessor 201 configured to execute applications for:

• • [0440] i. a first Analog to Digital converter 204,
  • [0441] ii. a second Analog to Digital converter 205,
  • [0442] iii. a first frequency analysis module 206, configured to extract a plurality of frequency components from the first audio signal,
  • [0443] iv. a second frequency analysis module 207, configured to extract a plurality of frequency components from the second audio signal,
  • [0444] v. a first power estimation block 208, configured to calculate the sound intensity in decibels of each frequency component of the first audio signal 002,
  • [0445] vi. a second power estimation block 209, configured to calculate the sound intensity of decibels of each frequency component of the second audio signal 003,
  • [0446] vii. a first dBFS to dBSPL Converter 210,
  • [0447] viii. a second dBFS to dBSPL Converter 211,
  • [0448] ix. a parameter estimation module, which determines parameter settings for a parallel compression module, according to information stored in memory,
  • [0449] x. the parallel compression module 215, parameterized by threshold T, compression ratio R, and makeup gain g0, where the parameters may by dynamically varied, where the parallel compression module is comprised of:
    • [0450] a. a linear compression module 214, which applies gains to the first audio signal according to the equation
      G=g0+(P−T)*(1/R−1),
    • [0451] where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
    • [0452] b. a summer, which combines the resulting amplified first audio signal component with the original first audio signal component,
  • [0453] xi. a Frequency Synthesis Module 216; and
  • [0454] xii. a Digital to Analog Converter 219;
    wherein (i) the first audio source 202 extracts the first audio signal x(t), where the first audio signal is then transmitted to the first Analog to Digital Converter 204 thereby converting the first audio signal into digital format; and, (ii) the second audio source 203 extracts the second audio signal x₀(t), where said second audio signal is then transmitted to the second Analog to Digital Converter 205 thereby converting the second audio signal into digital format,
    whereupon the first audio signal and the second audio signal are fed, respectively, through the First Frequency Analysis Module 206 and Second Frequency Analysis Module 207 breaking down each respective audio signal into separate components,
    whereupon, for each frequency component, the apparatus 200 identifies a source signal and a background signal corresponding to a frequency component 250, and said signals are thereafter transmitted to the microprocessor 201,
    whereupon, for each frequency component, both the source signal and the background signal are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the signals, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS 252,
    whereupon, for each frequency component, and at each instant, the microprocessor, using the power estimates of the source signal and background signal, obtains parameters for the parallel compression, depending on the sound intensity of the background signal at each instant,
    whereupon the microprocessor, applies the parameters to a parallel compression module,
    whereupon the parallel compression module applies a linear compression, according to the selected parameters, to each of the frequency components of the first audio signal,
    whereupon the parallel compression module sums each frequency component of the first audio signal with the corresponding compressed audio signal,
    whereupon the apparatus 200 reconstitutes the plurality of frequency components of the corrected first audio signal using the Frequency Synthesis Module 218, to produce a corrected first audio signal,
    whereupon the corrected first audio signal is fed through the Digital to Analog Converter 219, resulting in analog signal x_out(t).
    3. The apparatus of claim 1-2, where the First and Second Frequency Analysis Modules are composed of digital filter banks which are IIR filters that break the signal down into various sub-band channels x(t) and x₀(t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module is a summer operation that combines the various channels into output x_out(t).
    4. The apparatus of claim 3, where the various sub-band channels correspond to critical bands of hearing.
    5. The apparatus of claim 1-2, where the First and second Frequency Analysis Modules are digital filter banks composed of wavelet filters.

[0455] 6. The apparatus of claim 1-2, where the First and Second Frequency Analysis Modules are composed of FFT coprocessors that perform Fourier transforms on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module is an FFT coprocessor that performs an inverse Fourier transform, resulting in output signal x_out(t).

7. The apparatus of claim 6, where the frequency components are grouped together to approximate sub-bands correspond to critical bands of hearing.

[0456] 8. The apparatus of claim 1-2, where the psychoacoustic model 22 consists of a forward model which computes subjectively perceived Sones in terms of Phons, and an inverse model which computes Phons in terms of Sones, allowing the Psychoacoustic model to be combined into the Conductor Equation, the Conductor Equation 16 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,
where this equation may be stated as the desired gain in Phons being the Phons amplification necessary to amplify the sound over background noise so that it's subjectively perceived loudness over background noise is the same as the subjectively perceived loudness without the noise, where the function S(P) is found by solving the Earspring equation under experimental boundary conditions, where the Earspring equation 14,15 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), which is a harmonic oscillator model of the signal processing performed by the ear, the cochlea, and immediate processing by the auditory cortex, where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration for a given frequency channel, identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and r=the scale factor, where for a particular frequency of the driving force F(t), which represents a pure sinusoid of intensity P, a solution y(t) to this equation can be found, which can then be converted to <y²>, which represents the perceived loudness S in Sones, resulting in a function S(P),

[0457] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10},$
where S(P) is the perceived loudness in Sones which the listener experiences for a particular sub-band, and P is the sound intensity impinging on the ear in dBPhons for a particular subband, where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency of the earspring, where the constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions,
9. The apparatus of claim 8, where the set of boundary parameters used to obtain solutions to the Earspring Equation comprises: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz.
10. The apparatus of claim 9, where a computational model is used to estimate Earspring equation solutions S(P) and S⁻¹(S) consisting of quadratic functions such as

[0458] $S\left(P\right)\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring equation and are stored in long term memory.
11. The apparatus of claim 9, where the psychoacoustic model is replaced by a computational model 124 which is used to approximate the output of the combined psychoacoustic model and conductor equation, where the computation model comprises a quadratic function

[0459] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Conductor equation 116, for various values of the threshold elevation P₀ and are stored in a lookup table 123.
12. The apparatus of claim 9, where the psychoacoustic model 122 is replaced by any numerical function approximator 124 that is fit against the Earspring 114, 115 and Conductor equations 116, and approximates their solutions, where these approximators may include, but are not limited to linear regression, neural networks, polynomial regression, logistic regression, kernel regression, factor analysis, spline fitting, piecewise curve fitting, radial basis functions, and any other suitable approximator now known or later discovered.
13. The apparatus of claim 2, where discrete sets of parameters are found by dividing the range of possible noise volumes into discrete levels, and where sets of parameters are found for each discrete noise loudness level, where the parameter estimation module 111 is a lookup table, where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.
14. The apparatus of claim 2, where the parameter estimation module 111 is a polynomial continuous curve defining the relationship between the ambient noise loudness and each of the parameters, where polynomial regression is used to find parameters which define this curve, where these parameters are then stored in memory, so that during processing the parameters can then be computed using a polynomial equation.
15. The apparatus of claim 2, where one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.
16. The apparatus of claim 2, where the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
G=g0+(dP−T′)/R−dP
where dP is the difference between the signal loudness and noise loudness in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.
17. The apparatus of claim 16, where a fixed set of parameters is used.
18. The apparatus of claim 17, where the fixed parameters fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.
19. The apparatus 100 of claim 1-2, wherein the apparatus is a mobile phone.
20. A method of obtaining the solution to the Earspring equation 114,115 and it's inverse function in a computationally efficient manner, using a computational model in which various parameters of the model are fit against a variety of solutions of the Earspring equation, and it's inverse, for various input amplitudes and/or frequencies, where the parameters are stored within a long term memory component. In one exemplary embodiment, the computational model that estimates S(P) may consist of a quadratic function

[0460] $S\left(P\right)\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring equation and are stored in long term memory.
21. A method of obtaining corrective gains ΔP corresponding to a previously determined spectrum of threshold elevations, using the Conductor Equation 116 and perceived loudness models S(P) and S⁻¹(S) based on solutions to the Earspring equation, the Conductor Equation 116 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,
where S(P)=a Sones equation from an EarSpring equation 114,115 solution for the given frequency, as a function of Phons Pressure, and S⁻¹(S) is an inverse function of this solution that gives the sound intensity P in Phons needed for a normal listener to perceive the sound at loudness S in Sones, where P(f)=a Phons amplitude of the frequency component at f of a source signal of interest, and P₀(f)=a Phons amplitude of the frequency component at f of a background signal of interest, where Sones is a measure of perceived loudness under subjective experience and Phons is a measure of objective sound intensity, where the Earspring equation 114,115 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), which is a harmonic oscillator model of the signal processing performed by the ear, the cochlea, and immediate processing by the auditory cortex, where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration for a given frequency channel, identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor,
where for a particular frequency of the driving force F(t), which represents a pure sinusoid of intensity P, a solution y(t) to this equation can be found, which can then be converted to <y²>, which as the output of the equation represents the perceived loudness S in Sones, producing a function S(P),

[0461] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10},$
where S(P) is the perceived loudness in Sones which the listener experiences for a particular frequency, and P is the sound intensity impinging on the ear in dBPhons for a particular frequency, where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency of the earspring, where the constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions,
22. The method of claim 21, where the set of boundary parameters comprises: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz.
23. The method of claim 21, where the gains ΔP are obtained in a computationally efficient manner, using a single computational model 124 that combines S(P) and its inverse in the Conductor Equation 116 to obtain a single computational model

[0462] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the Conductor equation and are stored in a lookup table 123.
24. The method of claim 21, where the psychoacoustic model is a computational model 124 which may be any numerical function approximator that is fit against the Earspring and Conductor equation and approximates their solutions, where these approximators may include, but are not limited to linear regression, neural networks, polynomial regression, logistic regression, kernel regression, factor analysis, spline fitting, piecewise curve fitting, radial basis functions, and any other suitable approximator now known or later discovered.
25. A method of using parallel compression to approximate a desired non-linear dynamic range compression curve, wherein a compression curve defines the output loudness in decibels versus the input loudness in decibels of a signal of interest, wherein the compression is implemented by applying a gain G to the signal of interest according to the following equation:
G=g0+(P−T)*(1/R−1),
where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
whereupon the linearly compressed signal is added together with the original signal, producing a parallel compressed output signal, where the overall amplitude of the input versus the output signal defines a parallel compression curve, where g0, T, and R are thus parameters that vary the shape of the parallel compression curve, and P is the input loudness of the signal of interest, where the method is comprised of the following steps:

[0463] a) obtaining parameters that shape the parallel compression curve, by fitting compression curves using these parameters against a plurality of desired compression curves and selecting the parameters that minimize the difference between the compression curve of the parallel compression module and each of the desired compression curves,

[0464] b) storing the optimal sets of parameters in a memory device,

[0465] c) during real-time processing, at each time step:

• • [0466] i. retrieving a set of parameters, depending on the desired compression curve to be used,
  • [0467] ii. applying the set of parameters, to the parallel compression module; and processing a first audio signal using the time-varying compression parameters. in the parallel compression module, where the parallel compression module is performed by:
    • [0468] 1. applying linear compression to the first audio signal according to the equation G=g0+(P−T)*(1/R−1), producing a linearly compressed
    • [0469] 2. summing the linearly compressed audio signal with the first audio signal
      26. The method of claim 25, where the plurality of desired dynamic range compression curves approximate the ideal compression curves needed to compensate for a plurality of ambient noise signal intensities, so that the derived parallel compression parameters correspond to a plurality of ambient noise loudness levels.
      27. The method of claim 26, where discrete sets of parameters are found by dividing the range of possible noise volumes into discrete levels, and where sets of parameters are found for each discrete noise loudness level, where the sets of parallel compression parameters are stored in a lookup table, and where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.
      28. The method of claim 26, where polynomial regression is used to fit a continuous curve between the ambient noise loudness and each of the parameters, where the parameters of this curve are stored, and where during real time processing the ambient noise loudness is measured and applied to the polynomial function to obtain each parameter T, R, and g0.
      29. The method of claim 26, where one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.
      30. The method of claim 26, where the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
      G=g0+(dP−T′)/R−dP
      where dP is the difference between the signal loudness and noise loudness in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.
      31. The method of claim 30, where a fixed set of parameters is used.
      32. The method of claim 31, where the fixed parameters fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.
      33. The method of claim 26, wherein the signal is divided into a plurality of sub-bands and parallel compression is applied independently to each channel.
      34. The method of claim 25, wherein a threshold setting for the parallel compression is set according to a current threshold elevation value, wherein the threshold elevation is a function of a second audio signal.
      35. A computer-readable medium 670 storing a set of instructions executable by one or more processors, where the computer-readable medium 670 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of frequency components which constitute a first audio signal, wherein corrective gains are generated for each frequency component by solving for a correction value of ΔP for a source signal, wherein said source signal is a frequency component of the first audio signal corresponding to a particular frequency component, where ΔP is derived from solving for the output of a psychoacoustic model 615,
      the computer-readable medium comprising:

[0470] a) code for solving for the power P(f) in Phons of the source signal of the first audio signal, at each of the frequency components of the first audio signal:

• • [0471] i. code for extracting a power value of the source signal at a frequency component, in dBFS with a first power estimator block, comprising:
    • [0472] 1. extracting the first audio signal x(t) 602, said first audio signal comprising a plurality of frequency components, where said first audio signal is then transmitted to an Analog to Digital Converter 604 thereby converting the signal into digital format,
    • [0473] 2. feeding the first audio signal through a First Frequency Analysis Module 606 where the first audio signal is broken down into separate frequency components,
    • [0474] 3. identifying a the source signal of the first audio signal corresponding to a frequency component 650,
    • [0475] 4. generating a power value for the source signal 651, whereby said power value is then fed into a dB Converter to generate a power value P_dBFS(f) for the source signal in dBFS 652,
  • [0476] ii. code for calculating a calibration offset to convert the power value P_dBFS(f) of the source signal in dBFS to a power value P_dBSPL(f) in dBSPL 610 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,
  • [0477] iii. code for converting the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation 612, producing P(f),
    • [0478] b) code for solving for P₀(f) in Phons of the background signal, at the frequency component:
      • [0479] i. code for extracting a power value of the background signal at the frequency component, in dBFS with a second Power Estimator Block, comprising:
        • [0480] 1. extracting a second audio signal x₀(t) 603, said second audio signal comprising a plurality of frequency components, where said second audio signal is then transmitted to a second Analog to Digital Converter 605 thereby converting the signal into digital format,
        • [0481] 2. feeding the second audio signal through a Second Frequency Analysis Module 607 where the second audio signal is broken down into separate frequency components,
        • [0482] 3. identifying a particular background signal, corresponding to the frequency component 650,
        • [0483] 4. obtaining a power value for the background signal, 651 whereby said power value is then fed into a dB Converter to generate a power value P_0—dBFS(f) for the background signal in dBFS 652,
      • [0484] ii. code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dbSPL(f) in dBSPL using the calibration equation 611 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,
      • [0485] iii. code for converting the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation 613, producing P₀(f),
    • [0486] c) code for using the psychoacoustic model 616 to solve for ΔP(f) using P(f) and P₀(f);
    • [0487] d) code for calculating a correction gain of power G(f), for a frequency f, in dBSPL based on the value of ΔP(f) in dB Phons, using a Gain equation 617;
    • [0488] e) code for reconstituting the first audio signal with corrective gains to the source signals at each of the frequency components of the first audio signal, comprising
      • [0489] i. feeding the correction gain to an amplifier wherein the amplifier amplifies the source signal at the frequency component by the corrective gain G(f) 653,
      • [0490] ii. combining, using a Frequency Synthesis Module 618, the source signals of the corrected signal, producing a corrected first audio signal, x_out(t),
    • [0491] f) code for feeding the corrected first audio signal through a Digital to Analog Converter 619.
      36. A computer-readable medium 1070 storing a set of instructions executable by one or more processors, where the computer-readable medium 1070 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of frequency components which constitute a first audio signal, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the compression is implemented by applying a gain G to the signal of interest according to the following equation:
      G=g0+(P−T)*(1/R−1),
      where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
      whereupon the linearly compressed signal is added to the first audio signal, where the overall amplitude of the input versus the output signal defines a parallel compression curve, where g0, T, and R are thus parameters that vary the shape of the resulting compression curve, where the psychoacoustic model takes as inputs a source signal loudness and a noise signal loudness, where the model computes the gain in sound loudness at a given frequency component required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness level, are determined beforehand and stored in memory, where a different set of parameters is retrieved at each instant depending on the ambient noise signal intensity at each frequency component, where the dynamically changing noise signal intensity results in a dynamically changing selection of parameters for the parallel compression module,
      the computer-readable medium comprising:

[0492] a) code for, at each frequency component, extracting a power value of the source signal at a frequency component, in dBFS with a first power estimator block 1008; comprising:

• • [0493] i. extracting the first audio signal x(t) 1002, said first audio signal comprising a plurality of frequency components, where said first audio signal is then transmitted to an Analog to Digital Converter 1004 thereby converting the signal into digital format,
  • [0494] ii. feeding the first audio signal through a First Frequency Analysis Module 1006 where the first audio signal is broken down into separate frequency components,
  • [0495] iii. identifying a particular source signal of the first audio signal corresponding to a frequency component 1050,
  • [0496] iv. generating a power value for the source signal 1051, whereby said power value is then fed into a dB Converter to generate a power value P_dBFS(f) for the source signal in dBFS 1052,
  • [0497] v. code for calculating a calibration offset to convert the power value P_dBFS(f) of the source signal in dBFS to a power value P_dBSPL(f) in dBSPL 1010 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,

[0498] b) code for extracting a power value of the background signal at the frequency component, in dBFS with a second Power Estimator Block 1009; comprising:

• • [0499] i. extracting a second audio signal x₀(t) 1002, said second audio signal comprising a plurality of frequency components, where said second audio signal is then transmitted to a second Analog to Digital Converter 1004 thereby converting the signal into digital format,
  • [0500] ii. feeding the second audio signal through a Second Frequency Analysis Module 1006 where the second audio signal is broken down into separate frequency components,
  • [0501] iii. identifying a particular background signal, corresponding to the frequency component 1050,
  • [0502] iv. obtaining a power value for the background signal, whereby said power value is then fed into a dB Converter to generate a power value P_0—dBFS(f) for the background signal in dBFS 1051,
  • [0503] v. code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dbSPL(f) in dBSPL using the calibration equation 1010 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,

[0504] c) code for determining parameters for the parallel compression module 1022, including compression ratio, makeup gain, and threshold, based on the power value of the background signal, where the parameters have previously been fit against solutions to the Earspring and conductor equations, for a variety of background signal power values, and where the parameters may be retrieved at runtime, given a background signal power;

[0505] d) code for applying parallel compression to the source signal 1023, using the stored parameters now retrieved, thereby amplifying the source signal according to the desired compression

[0506] e) code for reconstituting the audio signal with the corrective gain to the source signal at the frequency component, comprising combining, using a Frequency Synthesis Module 1018, the parallel compressed frequency components of the audio signal, producing a corrected audio signal, xout(t).

[0507] 37. The computer readable media storing a set of instruction of claim 35-36, where the First and Second Frequency Analysis Modules are composed of Digital Filter Banks which are filters that break the signal down into various channels x(t) and x₀ (t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module is a summer operation that combines the various channels into output x_out(t).
38. The computer readable media storing a set of instruction of claim 35-36, where the First and second Frequency Analysis Modules are digital filter banks composed of wavelet filters.
39. The computer readable media storing a set of instruction of claim 37, where the various channels correspond to critical bands of hearing.
40. The computer readable media storing a set of instruction of claim 35-36, where the First and Second Frequency Analysis Modules perform a Fourier transform on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module performs an inverse Fourier transform, resulting in output signal x_out(t).
41. The computer readable media storing a set of instruction of claim 40, where the frequency components are grouped together to approximate sub-bands correspond to critical bands of hearing.
42. The computer readable media storing a set of instruction of claim 35-36, where the psychoacoustic model 122 consists of a forward model which computes subjectively perceived Sones in terms of Phons, and an inverse model which computes Phons in terms of Sones, allowing the Psychoacoustic model to be combined into the Conductor Equation, the Conductor Equation 116 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,
where this equation may be stated as the desired gain in Phons being the Phons amplification necessary to amplify the sound over background noise so that it's subjectively perceived loudness over background noise is the same as the subjectively perceived loudness without the noise, minus some correcting factor, where the function S(P) is found by solving the Earspring equation under experimental boundary conditions, where the Earspring equation 114,115 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), which is a harmonic oscillator model of the signal processing performed by the ear, the cochlea, and immediate processing by the auditory cortex, where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration for a given frequency channel, identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor, where for a particular frequency of the driving force F(t), which represents a pure sinusoid of intensity P, a solution y(t) to this equation can be found, which can then be converted to <y²>, which as the output of the equation represents the perceived loudness S in Sones, producing a function S(P),

[0508] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10},$
where S(P) is the perceived loudness in Sones which the listener experiences for a particular frequency, and P is the sound intensity impinging on the ear in dBPhons for a particular frequency, where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency of the earspring, where the constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions.
43. The computer readable media storing a set of instruction of claim 42, where the set of boundary parameters used to obtain solutions to the Earspring Equation comprises: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz.
44. The computer readable media storing a set of instruction of claim 35, where the computational model 122 is used to estimate S(P) and S⁻¹(S) consisting of a quadratic function

[0509] $S\left(P\right)\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring and Conductor equations and are stored in long term memory.
45. The computer readable media storing a set of instruction of claim 35, where the computational model 122 is used to estimate ΔP consists of a quadratic function

[0510] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring and Conductor equations and are stored in a lookup table 123.
46. The computer readable media storing a set of instruction of claim 35, where the computational model 122 is any numerical function approximator that is fit against the Earspring and Conductor equation and approximates their solutions, where these approximators may include, but are not limited to linear regression, neural networks, polynomial regression, logistic regression, kernel regression, factor analysis, spline fitting, piecewise curve fitting, radial basis functions, and any other suitable approximator now known or later discovered.
47. The computer-readable media storing a set of instructions of claims 35-36, wherein the first audio signal comprises a plurality of frequency components and wherein a corrective gain corresponding to each respective component is processed and applied to each component in parallel, before summation.
48. The computer-readable medium storing a set of instructions of claims 35-36, wherein the first audio signal contains more than one frequency component and steps (a) through (d) are repeated in series for all frequencies in the spectrum, before proceeding to step (e).
49. The computer-readable medium storing a set of instructions of claims 35-36, wherein the first audio signal contains more than one frequency component and steps (a) through (d) are in part computed in parallel and in part computed in series, before proceeding to step (e).
50. The computer readable medium storing a set of instructions of claims 35-36, in which either the Earspring equation or its inverse are estimated using separate computational models as in claim 5, replacing either step (a) (vi) or (a)(vii), respectively.
51. The apparatus of claim 1, and the computer readable medium storing a set of instructions of claim 35, wherein the dBSPL to dBPhon Conversion equation 12,13 is:

[0511] $P_{\mathrm{dBPhon}}\left(F\right)=\frac{P_{\mathrm{dBSPL}}\left(F\right)-\mathrm{ATH}\left(F\right)}{1+\mathrm{kATH}\left(F\right)}$

• • [0512] where
  • [0513] F=frequency
  • [0514] ATH(F)=Absolute Threshold of Hearing
  • [0515] K=− 1/120 when ATH(F)>0
    • [0516] =+ 1/240 when ATH(F)<0
      52. The apparatus of claim 1, and the computer readable medium storing a set of instructions of claim 35, wherein the Gain equation 17 is:
      ΔP_dBSPL(F)=ΔP_dBPhon(F)(1+kATH(F))
  • [0517] where
  • [0518] F=frequency
  • [0519] ATH(F)=Absolute Threshold of Hearing
  • [0520] K=− 1/120 when ATH(F)>0
    • [0521] =+ 1/240 when ATH(F)<0
      53. The apparatus of claim 1, and the computer readable medium storing a set of instructions of claim 35, where the absolute threshold of hearing is computed by the formula:

[0522] $\mathrm{ATH}\left(F\right)=\frac{3.64}{F_{\mathrm{KHz}}^{0.8}}-6.5e^{-0.6{\left(F_{\mathrm{kHz}}-3.3\right)}^2}+0.001F_{\mathrm{kHz}}^4-3.37$
54. The computer readable media storing a set of instruction of claim 36, where discrete sets of parameters are found by dividing the range of possible noise signal intensities into discrete levels, and where sets of parameters are found for each discrete noise intensity level, where the parameter estimation module 111 is a lookup table, where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.
55. The computer readable media storing a set of instruction of claim 54, where the parameter estimation module 111 is a polynomial continuous curve defining the relationship between the ambient noise loudness and each of the parameters, where polynomial regression is used to find parameters which define this curve, where these parameters are then stored in memory.
56. The computer readable media storing a set of instruction of claim 54, where one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.
57. The computer readable media storing a set of instruction of claim 54, where the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
G=g0+(dP−T′)/R−dP
where dP is the difference between the signal intensity and noise intensity in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.
58. The computer readable media storing a set of instruction of claim 57, where a fixed set of parameters is used.
59. The computer readable media storing a set of instruction of claim 58, where the fixed parameters fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.
60. The apparatus of claim 1-2, and the computer readable medium storing a set of instructions of claim 35-36, where there are a plurality of source audio signals corresponding to the various signals of a stereo or surround sound audio system, and where gains for each channel are independently computed and applied to each channel.
61. The apparatus of claim 1-2, and the computer readable medium storing a set of instructions of claim 35-36, where there are a plurality of source audio signals corresponding to the various signals of a stereo or surround sound audio system, and a plurality of background noise signals, one for each channel, and where gains for each channel are independently computed, according to the source and noise signal for each channel, and applied to each channel.
62. The apparatus of claim 1-2, where the microprocessor is a DSP chip.

[0523] The present invention relates to the field of audio processing and more specifically to compensation for noise in the listener's environment.

[0524] A headset may be a set of headphones for allowing the user to listen to audio signals, or may be a two-way communication device worn on the head with a microphone for the user to speak through.

[0525] Headsets are often worn by personnel in environments of varying noise. Noise cancellation is a common technique used to compensate for noise, to allow the user to hear what is sent through the headset without the need to constantly adjust the volume. However, noise cancellation, by definition, reduces the user's awareness of external noises, which depending on the user's environment, he may need to be aware of. For example, military personnel working in noisy combat environments need to hear commands sent over headphones, but also need to be aware of sounds such as gunfire in their vicinity to maintain situational awareness. A person listening to music while crossing a busy street similar may need to be aware of approaching vehicles.

[0526] For these reasons, noise cancellation often has the undesirable effect of making people unaware of approaching danger. Hence there is a need to compensate for noise in other ways. Some headsets may use automatic level control to adjust the volume of signals automatically depending on the total volume of noise. However, because of the non-linearities and dynamic properties of the human ear, this technique often amplifies some sounds too much and other sounds too little.

[0527] There are many systems for the application of filtering to noise suppression in an audio signal. In most cases, these inventions relate to the removal of noise that is present in the source signal from the origin or introduced into the signal through processing and transmission. Various forms of filtering may be applied which suppresses the noise signal in whole or in part. Generally, these systems have adverse impacts upon the quality of the original signal. However, these systems do not address noise in the environment of the listener, which cannot be filtered.

[0528] Conversely, systems for the suppression of noise in the listener's environment also exist. These systems generally use noise cancellation to remove the disrupting external signal by adding sound projected through headphones which has the effect of countering the sound waves produced by the noise. In this case, the noise is completely canceled and listener is generally unaware of the existence of the external noise, a result which can reduce the awareness of the listener to potential dangers in the environment.

[0529] In some prior art systems, dynamic volume compensation may be used to raise the volume of a source signal of interest over ambient background noise. However, these systems increase the gain in a spectrally uniform manner, raising the volume of all frequency components equally. This effect can distort the perception of music and speech due to the non-linear behavior of the human ear with respect to frequency and volume.

[0530] Microphones and mechanical systems (e.g., computer software) can measure dBSPL; a sound (e.g., 40 dBSPL) at particular frequency (e.g., 1 kHz) sounds just as loud as the sound (e.g., 40 dBSPL) at a different frequency (e.g., 4 kHz) to a microphone or mechanical system. However, our hearing can be affected by the mechanical construction of our outer ear and/or slow variation in sensitivity across the basilar membrane due to fluid damping of the incident waves in the cochlear fluid. The variable sensitivity of human hearing is reflected in the Fletcher-Munson equal loudness contours and the equal-loudness contours from ISO 226:3003 revision (Phons). The equations of the systems of the present invention utilize conversions from dBSPL to Phons and from Phons to dBSPL (incoming sounds levels are converted from dBSPL to Phons for use in the equations, then subsequently the Phons are converted to dBSPL for expression to speakers and headphones. Conversion from dBSPL to Phons and from Phons to dBSPL is in accordance with the Fletcher-Munson equal-loudness contours and the equal-loudness contours from ISO 226:3003 revision.

[0531] Since the human ear dynamically adjusts to sound intensity levels, the presence of background noise alters the threshold at which sounds begin to be perceived. As a result, ambient noise at a given frequency may make sounds at those frequencies that would otherwise be perceptible imperceptible in the presence of ambient noise. In order for the sound to be heard it must be amplified over the background noise. The volume of the ambient noise therefore represents a degree of hearing impairment or baseline threshold elevation over which the sound must be amplified to be perceived.

[0532] This effect varies according to the spectral composition of the noise, that is, spectral components that are sufficiently far from the spectral composition of the noise will remain perceptible. Consequently, using the total intensity of the background noise to raise the intensity of the source uniformly will overly amplify bands which are not affected, possibly raising the volume to damaging levels. In order to amplify only those components which need compensation, the gains to the source signal must vary by spectral band, according to the spectral composition of the noise.

[0533] Moreover, due to the nonlinear response of the human ear, using the spectral intensity of the background noise at a particular band as the gain for the source at that band will produce excessive amplification. In order to compute the correct gain a nonlinear corrective model must be used to compute an appropriate gain for each frequency. The intensity of the background noise as well as the source signal at a given frequency are inputs to this model, and the output is a desired gain for the source signal at that frequency.

[0534] The perceived loudness of sound can be modeled by a harmonic oscillator with spring constant that varies according to the mean power of vibration. This model is called the Earspring model. The Earspring equation 10,11 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration or S(P), identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor. Thus the resonant frequency of the Earspring varies with the amplitude of the <y²> term. Since <y²> is a function of y, this equation is nonlinear.

[0535] For a particular driving force F(t), which we will consider to be a sinusoid at a given frequency and amplitude, for which the amplitude is P in dBPhons, a solution to the earspring equation y(t) can be found, for a particular set of boundary conditions, which is the steady state response to the forcing function. Transforming into the frequency domain we can obtain the mean power of vibration <y²>=½|Y|², which can be rewritten as S(P).

[0536] A possible set of boundary parameters comprises (i) a Sones ratio between the hearing threshold level and the reference level of 40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz; This equation can be solved for numerically though in practice it is more efficient to use a computational model to estimate the solution.

[0537] We can derive an equation for S(P),

[0538] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10}$
where S(P) is the perceived loudness in Sones which the listener experiences, and P is the sound intensity impinging on the ear in dBPhons. Where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency for a ˜40 dBSPL amplitude driving force. The constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions. Note that this formula is independent of frequency although the sound intensity P of any given signal may vary as a function of frequency.

[0539] The present invention features systems for dynamically adjusting audio signals by applying a gain to the signal in a spectrally varying manner to compensate for a noise source in the environment of the listener, such that the sound is perceived by the listener to be unchanged in loudness and spectral composition. The system obtains a threshold elevation for each frequency component by analyzing the spectral composition of the ambient noise. This threshold elevation is then used by a perceptual model of hearing to determine an appropriate gain adjustment for the corresponding frequency component of the source signal which will make that component perceived by the human ear to be just as loud as if the noise were not present. After applying the gains to the source signal, the resulting signal is output to the speaker. The system allows a listener to hear without distortion, over ambient noise, by applying a gain to the source that varies according to the spectral composition of the noise, rather than cancelling the noise, or applying a uniform volume adjustment to the source. The perceived spectral composition of the source is thus adjusted without the removal of the noise signal. Systems may be incorporated into apparatuses including but not limited to headsets and music players.

[0540] Any feature or combination of features described herein are included within the scope of the present invention provided that the features included in any such combination are not mutually inconsistent as will be apparent from the context, this specification, and the knowledge of one of ordinary skill in the art. Additional advantages and aspects of the present invention are apparent in the following detailed description.

The Following Information Regarding Terms is Non-Limiting and Exemplary in Nature for the Purpose of Understanding the Spirit of the Invention

• • [0541] 1. Amplifier: This component can be any amplifier or related device capable of applying gains to an audio signal.
  • [0542] 2. Audio Codec: Audio Encoder/Decoder converts an audio signal to and from its encoded format either for storing in a file or transmission over a network
  • [0543] 3. A/D Converter: Converts audio signals from analog to digital format. In some embodiments, the term A/D Converter is used interchangeably with a means for converting an audio signal from analog to digital form. Analog-to-digital and digital-to-analog converters can be assumed to exist at interface points between analog elements and digital elements of any embodiment. Said means may be electronic, non-electronic or only partially electronic. Said converters, also commonly referred to as Analog to Digital Converters or ADCS, are well known within the art and would be understood by one skilled in the same. Said converters could include but are not limited to direct-conversion ADCs, successive approximation ADCs, a ramp-compare ADC, Wilkinson ADCs, integrating ADCs, Delta-encoded ADCs, pipeline ADCs, time interleaved ADCs, or ADCs with intermediate FM Stage
  • [0544] 4. Audio Device: As used herein, an audio device is any audio device capable of extracting, storing or receiving an audio signal. In some embodiments, the term audio device is used interchangeably with a “means for extracting and/or identifying an audio signal of interest and/or ambient noise signal.” Said means may be, for example, a microphone, receiver, or storage media containing a signal of interest in an audio file and code for reading the signal from an audio file.
  • [0545] 5. Audiogram: One form of defining a person's hearing loss which plots thresholds of hearing relative to a standardized curve that represents ‘normal’ hearing, in dB(HL). An audiogram can be obtained using a behavioral hearing test called Audiometry. The “Tone” test involves presenting different tones at a specific frequency (pitch) and intensity (loudness) and determining thresholds of hearing for each tone where the thresholds correspond to how loud the tone must be presented in order for the person to perceive the tone sound.
  • [0546] 6. Audio Jack: An input by which an audio signal can be received. i.e. from a microphone.
  • [0547] 7. Background signal: The background signal, as used herein, refers to a portion of the second audio signal or the ambient noise signal at the same frequency component as that of the source signal. The source signal and the background signal are paired.
  • [0548] 8. Calibration equation: Calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed. It is represented by: P_dbSPL(F)=P_dBFS(F)+(P_dBSPL—0(F)−P_dBFS—0(F)) wherein PdBSPL0 is nominally a value between 65 and 83, and wherein P_dBFS0 is nominally a value between −20 and −12;
    • [0549] where
      • [0550] P=Phons amplitude of a source signal of interest
      • [0551] F=Frequency
      • [0552] dBSPL=decibels according to sound pressure
      • [0553] dBFS=decibels relative to full scale
  • [0554] 9. Bark frequency: The bark scale is a psychoacoustic scale for subjective measurements of loudness. The scale is broken down into critical bands of hearing. Critical bands are known frequency bandwidths that correspond to limitations of the human ear to perceive sounds with frequency differences smaller than the critical bandwidths.

[0555] 10. D/A Converter: Converts audio signals from digital to analog format. In some embodiments, the term D/A Converter is used interchangeably with a means for converting an audio signal from digital to analog form. Analog-to-digital and digital-to-analog converters can be assumed to exist at interface points between analog elements and digital elements of any embodiment. A D/A converter, or DAC, is capable of converting a digital, usually binary signal code to an analog signal (current voltage or electric charge). DACs may include but are not limited to pulse width modulators, oversampling or interpolating DACs, binary weighted DACs, R-2R Ladder DACs, Successive approximation or cyclic DACs, thermometer coded DACs, and hybrid DACs.

• • [0556] 11. Digital Signal Processor (“DSP”) Chip: A specialized microprocessor with an architecture for the fast operational needs of digital signal processing.
  • [0557] 12. FFT Co-Processor: A specialized microprocessor designed for the purpose of rapidly computing Fourier transforms using a algorithm for computing Fourier transforms commonly known as the ‘Fast Fourier Transform’ (FFT).
  • [0558] 13. Field Programmable Gate Array (“FPGA”): a reprogrammable electronic chip.
  • [0559] 14. Filter banks: An array of band-pass filters that separate an input signal into multiple components, or as used herein, frequency components, each frequency component carrying a single frequency subband of the original signal.
  • [0560] 15. FIR Filter: A filter with an impulse response (or response to any finite length input) of finite duration, because it settles to zero in finite time. This is in contrast to infinite impulse response (IIR) filters, which have internal feedback and may continue to respond indefinitely (usually decaying). FIR filters can be discrete-time or continuous-time, and digital or analog.
  • [0561] 16. Fourier transforms: A mathematical operation that computes the frequency content of an audio signal, taking a given audio signal as input and outputting magnitudes and phases across the frequency spectrum.
  • [0562] 17. Frequency Analysis Model: This model can use a variety of known methods to divide the audio signal of interest into a plurality of frequency components. Such methods may include, but are not limited to methods such as filter banks, Fourier transforms, wavelet transforms or other signal processing techniques. In some embodiments, the term Frequency Analysis Model is used interchangeably with a “means for dividing an audio signal into a plurality of frequency components. Said means may be, for example, a digital filter bank, analog filter bank, FFT Co-processor, or code for computing a Fourier transform analysis to an audio signal on computer-readable media.
  • [0563] 18. Frequency component and Frequency Component Spectrum: A frequency component, as used herein, identifies a portion of the frequency range of an audio signal. A frequency component can may comprises a particular, individual frequency, frequency channels, or frequency bands. A frequency component spectrum is a plurality of frequency components.
  • [0564] 19. Frequency Synthesis Module: A module which the present invention can use to reconstitute the various frequency components of a particular audio signal of interest. The frequency synthesis Module may perform summation of the various frequency components of the signal of interest, or may perform an inverse Fourier transform, or other transform, after said frequency components have been adjusted, to create a new signal or waveform, depending on whether the Frequency analysis module was a filter bank or a fourier transform, or other transform. “Sound Systems: Design and Optimization: Modern Techniques and Tools for Sound System Design and Alignment”, McCarthy, Bob (2010) (Focal Press) Second Edition. In some embodiments, the term Frequency Synthesis Module is used interchangeably with a “means for reconstituting an audio signal”. Said means may be, for example, a summer block, a frequency synthesis module, an FFT-Co processor, an inverse FFT transform, or code for implementing an inverse FFT transform on computer-readable media.
  • [0565] 20. Gain Block component—This component applies the gains to each corresponding frequency component or band. In some embodiments, the term gain block component is used interchangeably with a “means for applying gains to a audio signal.” Said means may be, for example, an amplifier, or gain block component.

[0566] 21. Infinite Impulse Response (“IIR”) Filter: A signal processing filter with an impulse response function that is non-zero over an infinite length of time. May be implemented as either digital or analog IIR filters. IIR filters use fewer computing resources because they use fewer taps.

• • [0567] 22. Input Device: An device configured to enable a user to set a parameter b of a personalized sound adjustment model. In some embodiments, the term input device is used interchangeably with “a means for enabling a user to set a parameter b of a personalized sound adjustment model”. Said means may be, for example, a physical input device such as a user adjustable input, a knob, a slider, or any combination thereof. Said means may alternatively be an electronic means or a combination of physical hardware and electronics such as graphical user interface. Said means may further be a touch screen configured to accept a user's choice of parameter b. Additional means would be recognized and understood by one skilled in the relevant art.
  • [0568] 23. Inverse Fourier Transform: A mathematical operation which computes a time domain signal corresponding to a given frequency spectrum.
  • [0569] 24. kHz frequency to critical band center frequency converter: The critical bands of hearing are natural bands in the audio frequency spectrum in the human ear. These bands are characterized by the masking of tones by other tones in the same critical band. There are several experimentally derived frequency scales that attempt to approximate the critical bands including the Bark scale—a psychoacoustic model proposed by Eberhard Zwicker in 1961, and named after Heinrich Barkhausen who proposed the first subjective measurements of loudness. One example of the equation for converting from kHz to critical band frequency is the conversion to Bark frequency:
    CBR(F_kHz)=26.811(1+1.960/F_kHz)−0.53
  • [0570]  In some embodiments, the term kHz frequency to critical band center frequency converter is used interchangeably with a “means for converting a frequency component or series of frequency components to critical band frequency. Said mean may be, for example, a summer.
  • [0571] 25. Linear slope equation: This equation is part of the Personalized sound adjustment model, and is a novel mathematical equation, and one potential embodiment by which threshold elevations are generated for vTuning. In some embodiments, the term Linear Slope equation is used interchangeably with a “means for estimating threshold elevations specific to a particular person.” Said means may be, for example, the Linear Slope equation, an audiogram, or a pattern library.
  • [0572] 26. Lookup Table: A tabular list of parameters indexed by some given variable, such as signal power. Used by the model to select a parameter or set of parameters depending on the current real-time value of a variable. In some embodiments, the term Lookup Table is used interchangeably with a “means for determining threshold elevations by using a lookup table.” Said means may be, for example, computer-readable media code for retrieving a threshold elevation value from a memory device using a user adjustable parameter b and frequency as indexes.
  • [0573] 27. Microphone: The microphone can be any microphone or the like capable of receiving an audio signal.
  • [0574] 28. Parallel Compression: applying a linear gain to a signal which amplifies softer sounds and subsequently adding this amplified sound back in to the original signal. Also known as New York compression. In some embodiments, the term parallel compression is used interchangeably with a “means for computing and/or applying corrective gains to an audio signal where said gains correspond to threshold elevations, using a combination of parallel compression and known psychoacoustic models” or a “means for using parallel compression to approximate a desired non-linear dynamic range compression curve.” Said means can for example, be any known psychoacoustic model such as the model disclosed herein. Said means may for example, be a model where a compression curve defines the output loudness in decibels versus the input loudness in decibels of a signal of interest, wherein the compression is implemented by applying a gain G to the signal of interest according to the following equation:
    G=g0+(P−T)*(1/R−1)
  • [0575]  where G is the gain in dB to be applied to a frequency component of the signal of interest, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of linear compression, whereupon the linearly compressed signal is added together with the original signal of interest, producing a parallel compressed output signal, where the overall amplitude of the input versus the output signal defines a parallel compression curve, where g0, T, and R are thus parameters that vary the shape of the parallel compression curve, and P is the input loudness of the signal of interest. In some embodiments, the term parallel compression is used interchangeably with a “means for applying parallel compression to a signal of interest.” Said means may be, for example, a parallel compression chip or combining a compressed signal with an original signal. In some embodiments, parallel compression may be a “means for applying linear compression to a signal.” Said means may for example be an amplifier. In some embodiments, the term parallel compression is used interchangeably with a “means for controlling the shape of a parallel compression curve.” Said means may be, for example threshold elevations, makeup gains, and compression ratios. In some embodiments, the term parallel compression is used interchangeably with a “means for finding the parameters which best fit a parallel compression curve against a psychoacoustic model”. Said means may be, for example, linear regression, least squares fit, and function approximation. In some embodiments, the term parallel compression is used interchangeably with a “means for storing parameters”. Said means may be, for example, use of a lookup table, flash memory, hard disk, and long-term memory. In some embodiments, the term parallel compression is used interchangeably with a “means for retrieving the parameters to be applied to a parallel compression module.” Said means may be, for example, a lookup table, polynomials, and a parameter estimation module.
  • [0576] 29. Parameter: The parameter is a single mathematical parameter value that when input to the Personalized Sound adjustment, generates may generate a spectrum of threshold elevations.
  • [0577] 30. Pattern Library: A numerical library composed of points and values which measure threshold elevations. In some embodiments, the term Pattern Library is used interchangeably with a “means for creating a pattern library for storing threshold elevations.” Said means may be, for example, obtaining sample threshold elevation patterns for varying frequency for various typical users (users indexed by b) or computer-readable media code for storing a set of patterns in memory indexed by a user adjustable parameter b and a frequency.
  • [0578] 31. Personalized sound adjustment model: A novel mathematical model which enhances sound through a process of non-linear compression. Bark frequency is one of several types of measures of the natural bandwidth of human hearing referred to as critical bandwidth. Critical bands are defined by the distances in frequency in which masking of tones occurs in human ear, roughly correspond to linear position on the cochlea. Thus, the present invention utilizes an equation that, using a single parameter, chosen by a user, generates a spectrum of threshold elevations that generally fall along a straight line when plotted on models using dBHL against Bark frequency where said threshold elevations can be used to adjust a sound signal for persons with hearing loss or persons who simply wish to enhance the audio signal to their preference. In some embodiments, the term personalized sound adjustment model is used interchangeably with “a means for executing a personalized sound adjustment model”. Said means may be computing the model on a microprocessor, a digital signal processor, or using code for executing the model on computer-readable media. In some embodiments, the term personalized sound adjustment model is used interchangeably with a “means for determining threshold elevations from ambient noise.” Said means may be, for example, dividing an ambient noise signal into frequency components, estimating the ambient noise power of a frequency component, and determining threshold elevations corresponding to a frequency component.
  • [0579] 32. Power Estimation Model/Block: estimating the power value of an audio signal. In some embodiments, the term Power Estimation Block/Model is used interchangeably with a “means for estimating a power value of an audio signal.” Audio power estimation can be conducted using a variety of known techniques which would be understood by one skilled in the art. For example, some use a Minimum Statistics approach which is based on tracking minima of a short term power estimate of the audio signal in bands, over time. If a Fourier transform of the signal is used the power estimate is the square of the magnitude of the frequency component.
  • [0580] 33. Psychoacoustic model: Any appropriate psychoacoustic models can be used by the present invention to compute gains needed to amplify sound to overcome the effects of ambient noise on sound perception or gains needed to account for preferred threshold elevations computed for users' of normal hearing. One example of a known psychoacoustic model is as follows:

[0581] $\frac{\sqrt[\alpha ]{P_{\mathrm{SIG}}^{\alpha }+P_{\mathrm{NOISE}}^{\alpha }-P_{\mathrm{THRQ}}^{\alpha }}}{P_{\mathrm{SIG}}}$

• • [0582]  where G is the gain ratio, PSIG is the signal intensity at a frequency in units of power, PNOISE is the signal intensity of the background noise, PTHRQ is the absolute threshold of hearing, and α=0.2 is a constant. In some embodiments, the term psychoacoustic model is used interchangeably with a “means for generating or computing corrective gains corresponding to predetermined threshold elevations”. Said means may be, for example, additional known psychoacoustic models, perceptual models, computational models, and models utilizing parallel compression fit against known psychoacoustic models.
  • [0583] 34. RF Transceiver: Radio Frequency Transmitter/Receiver—this device interfaces directly with the antenna and can modulate an audio signal with its radio carrier frequency for transmission as well as demodulate a received radio frequency signal
  • [0584] 35. Source signal: The Source signal, as used herein refers to a portion of the first audio signal or the audio signal of interest at a given frequency component.
  • [0585] 36. Speaker: The speaker can be any speaker or the like capable of receiving and projecting a sound signal adjusted or corrected by the present invention.
  • [0586] 37. Storage media: A computer readable media used for storing data, such as a hard disk, RAM, or flash memory
  • [0587] 38. Summer block: Summation occurs when two or more audio signals are combined together to create a new wave form. Summation is the combining of two audio signals at the same frequency—not to be confused with mixing, which involves the combining of audio signals at different frequencies. Electrical summation occurs inside an electrical circuit. In some embodiments, the term summer block is used interchangeably with a “means for combining an audio signal with another audio signal”. Said means may be, for example, a summer.
  • [0588] 39. Threshold elevation: Threshold elevations correlate to the minimum sound pressures required to perceive sounds at various frequencies.
    Additional Terms:
  • [0589] 12. Conductor Equation (Perceptual Model Component): This is a novel equation which computes the gain necessary to amplify sound sufficiently to make the subjectively perceived sound appear to be as loud, to a hearing impaired subject, as if the subject had no hearing loss. In some embodiments, the term Conductor equation is used interchangeably with a means for determining the gain necessary to make a sound perceived to be as loud as if a threshold elevation were not present.
  • [0590] 13. dBSPL: Decibels sound power level—a unit of measure of sound intensity at the ear
  • [0591] 14. dBSPL to dBPhon Conversion equation: This equation converts sound pressure level to Phons in decibels:

[0592] $P_{\mathrm{dBPhon}}\left(F\right)=\frac{P_{\mathrm{dBSPL}}\left(F\right)-\mathrm{ATH}\left(F\right)}{1+\mathrm{kATH}\left(F\right)}$

• • [0593] where
    • [0594] F=Frequency
    • [0595] ATH=Absolute Threshold of Hearing
    • [0596] P=Phons amplitude of a source signal of interest
    • [0597] dBSPL=decibels according to sound pressure
  • [0598] 15. dBFS to dBSPL Converter: This equation converts the audio signal power from audio power in decibels to audio sound pressure level in decibels
  • [0599] 16. Ear Spring Equation/Crescendo (Psychoacoustic Model): This is a novel equation which computes the apparent subjective loudness for a normal subject, of a sound in Sones as a function of the input sound intensity of the sound in Phons. In some embodiments, the term “Ear Spring Equation/Crescendo” is used interchangeably with a “means for computing the loudness in Sones from the intensity in Phons at a person ear.” Said means may be the Earspring Equation and may also be any known psychoacoustic models.
  • [0600] 17. Gain Equation: This equation converts the gain in Phons to a gain in sound pressure level
    ΔP_dBSPL(F)=ΔP_dBPhon(F)(1+kATH(F))
  • [0601]  where:
    • [0602] F=Frequency
    • [0603] ATH=Absolute Threshold of Hearing
    • [0604] P=Phons amplitude of a source signal of interest
    • [0605] dBSPL=decibels according to sound pressure
  • [0606] 18. Output Device: A device configured to output an audio signal. In some embodiments, the term Output Device is used interchangeably with Audio Device. In some embodiments, the term Output Device is used interchangeably with a “mean for outputting an audio signal.” Said means may be, for example, a speaker, a transmitter, computer-readable media code for transmitting an audio signal or audio file and computer-readable media code for writing an audio signal to an audio file.
  • [0607] 19. Phons: Phons are objective units of sound intensity impinging upon the ear. Sound power may be converted from decibels (dBFS) to Phons using a conversion first to dBSPL then to Phons
  • [0608] 20. Psychoacoustic model (Crescendo): This is a novel equation derived from a complex combination of physics and mathematical equations, which computes gains needed to compensate for ambient noise using two major input values, a source signal of interest, and an ambient noise signal. In some embodiments, the term psychoacoustic model is used interchangeably with a “means for determining the gains to be applied to a signal of interest.” Said means may be, for example, a known psychoacoustic model or Crescendo.
  • [0609] 21. Sones: Sones are subjective units of perceived sound loudness
  • [0610] 22. Square Summer Equation: This equation squares the audio signal and adds the signal up N times, whereby the sum of the signal at the end block, divided by N, generates a mean power value for the audio signal. In some embodiments, this term is used interchangeably with a “means for extracting power values of an ambient noise signal.”

[0611] FIG. 1. Is a drawing of a possible systems architecture for a wired headset utilizing a DSP chip to implement an ambient noise correction compression algorithm.

[0612] FIG. 2. Is a drawing of a possible systems architecture for a wireless headset utilizing a DSP chip to implement an ambient noise correction compression algorithm.

[0613] FIG. 3. is a drawing of a possible architecture of the present invention implemented on a DSP microprocessor.

[0614] FIG. 4. Diagrams an implementation of the Perceptual Model of the present invention using analytical solutions to the Earspring model.

[0615] FIG. 5. Diagrams an implementation of the Perceptual Model in which a Numerical approximation to the solutions is made using a function approximator for which parameters have been stored in memory.

[0616] FIG. 6. Is a drawing illustrating an application of parallel compression to approximate a psychoacoustic non-linear compressor, where the parameters of the parallel compression are dynamically adjusted.

[0617] FIG. 7 is a drawing of an exemplary embodiment of the present invention, using a dynamically adjusted parallel compression to approximate the desired non-linear compression of the psychoacoustic model.

[0618] FIG. 8 is a plot of the compression ratio needed to make the parallel compression module approximate the psychoacoustic model, as found by fitting the parallel compression curves against the desired model curves.

[0619] FIG. 9 is a plot of the makeup gain needed to make the parallel compression module approximate the psychoacoustic model, as found by fitting the parallel compression curves against the desired model curves.

[0620] As shown in FIG. 1-3, in some broad embodiments, the present invention may comprise a headset 01, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment,

the headset 01 comprising:

• • [0621] a) an RF transmitter/receiver or an audio jack 113, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest
  • [0622] b) a speaker 110, coupled to the DSP chip 103, disposed upon the phone in a location ergonomically suitable for listening through the user's ear when using the phone,
  • [0623] c) a microphone 111, capable of receiving a second audio signal, which is noise, disposed upon the phone in a location removed from the locations of the first microphone 109, and speaker 110 suitable to minimizing the effect of the speaker's voice on the ambient noise measurement,
  • [0624] d) a second digital signal processing (DSP) chip, coupled to the audio codec, the microphone and the speaker, configured to implement a noise correction algorithm 103, the noise correction algorithm comprising:
    • [0625] i. a first Analog to Digital converter 204;
    • [0626] ii. a second Analog to Digital converter 205;
    • [0627] iii. a first Frequency Analysis Module 206;
    • [0628] iv. a second Frequency Analysis Module 207;
    • [0629] v. a first Power Estimator Block 208;
    • [0630] vi. a second Power Estimator Block 209;
    • [0631] vii. a first dBFS to dBSPL Converter 210;
    • [0632] viii. a second dBFS to dBSPL Converter 211;
    • [0633] ix. a first dBSPL to dB Phon Converter 212;
    • [0634] x. a second dBSPL to dB Phon Converter 213;
    • [0635] xi. a Perceptual model 222;
    • [0636] xii. a Gain Block component 217;
    • [0637] xiii. a Frequency Synthesis Module 218; and
    • [0638] xiv. a Digital to Analog Converter 219;
      wherein the DSP 103 receives a first audio signal which is the signal of interest via the RF transmitter 104,
      whereupon the DSP Chip 103, obtains a second audio signal from the background noise microphone 111;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First D/A Converter 204 and Second D/A 205 thereby converting the signals into digital format;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 206 and Second Frequency Analysis Module 207 breaking down each respective audio signal into arrays of frequency components;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS;
      whereupon (i) the DSP chip 103 completes a calibration offset such that a first dBFS to dBSPL Converter 210 converts the power value of the source signal from dBFS to dBSPL, and (ii) the DSP chip 103 completes a calibration offset such that a second dBFS to dBSPL Converter 211 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f);
      whereupon (i) a first dBSPL-dB Phon Converter 212 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 213 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f);
      whereupon the DSP chip 103 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the perceptual model 222 whereupon the DSP chip 103 calculates ΔP(f) in dBPhons
      whereupon the DSP chip 103 calculates a correction gain of power to the source signal G(f), at the frequency, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 217 which utilizes the Gain equation;
      whereupon the DSP chip 103 reconstitutes the first audio signal with the corrective gain to the source signal corresponding to the frequency, comprising:
  • [0639] a) multiplying the source signal at the frequency according to the corresponding corrective gain ΔP(f);
  • [0640] b) using the Frequency Synthesis Module 18, to reconstitute the resulting audio signal
    whereupon the corrected audio signal is fed through the Digital to Analog Converter 219, resulting in analog signal x_out(t);
    whereupon the DSP chip 103 outputs the corrected signal to the speaker 110.

[0641] As shown in FIG. 4, in some embodiments the perceptual model involves analytically computing the value in Sones of the input signal at a given frequency S(P), along with the estimated hearing loss due to background noise in Sones at that frequency S(P₀), and applying these values to the conductor equation to derive a gain in Phons for the signal that will correct for the perceived hearing loss, the Conductor Equation 13 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P

[0642] As shown in FIG. 5, in some embodiments the perceptual model uses a numerical estimate of these equations, combining the conductor equation and the Earspring equation solutions into one equation. In this embodiment, a function approximator, such as a quadratic estimator, linear regression with quadratic features, neural network, or other computational model, can be fit against the solutions to the Earspring equation, and the parameters of the model then stored in memory. During operation, these parameters can be retrieved by the microprocessor from a lookup table, and used to compute an approximate value for ΔP for any given P and P_0.

[0643] In some embodiments, all or part of the present invention could be implemented using a parallel processing architecture such that some or all of the set of computations from step 5 to 10 above may be computed simultaneously using multiple processing units.

[0644] In some embodiments, the absolute threshold of hearing may be computed in one example by the formula:

[0645] $\mathrm{ATH}\left(F\right)=\frac{3.64}{F_{\mathrm{KHz}}^{0.8}}-6.5e^{-0.6{\left(F_{\mathrm{kHz}}-3.3\right)}^2}+0.001F_{\mathrm{kHz}}^4-3.37$

[0646] In some embodiments, a Fourier transform of the signal is used so that the gain computation is performed for each element in the frequency spectrum instead of by channel.

[0647] In some embodiments, a computational model can be used in place of explicit computation of the Earspring and Conductor equations, where the computational model is

[0648] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1}$
where A, B, C, D and E are parameters which have been fit against solutions to the combined Earspring and Conductor Equations, for each value of P₀ in 5 dB increments, and stored in a lookup table, such that for a given P₀, parameters for A, B, C, D, and E are extracted from the table, and then the computational model is used to compute ΔP given P.

[0649] In some embodiments it is also possible to use separate computational models for S(P) and for S⁻¹(S), with separate sets of parameters fit against the solutions for various boundary parameters and values of P_0.

[0650] As shown in FIG. 6, in some embodiments, an alternative means of effectuating the gains demanded by the psychoacoustic model is to harness an existing sound processing technique known as parallel compression. Parallel compression involves compressing a copy of an original signal using linear compression 312, and subsequently adding this compressed signal back into the original signal. This results in a non-linear compression curve that amplifies low sounds below a certain threshold. The shape of the compression curve is controlled by a number of parameters including the threshold (T), compression ratio (R), and makeup gain (M). By varying these parameters, the parallel compression curves can be made to approximate the compression curve that would otherwise be produced by applying the psychoacoustic model. This is done by fitting the parameters against the desired characteristics of the psychoacoustic model using a least-squares error fit. These parameters are then stored in memory in a lookup table 311, or computed using a formula and may be retrieved in real time to dynamically update the parallel compression block 312. Since the psychoacoustic model depends upon both the signal and noise power, both signal and noise power 306, 307 will factor into determining the gains. Thus, these parameters must be updated at each time step account for the changing noise. Typically the T parameter is chosen to be the threshold elevation (or noise power) minus 20 dB. From that point makeup gain and compression ratio are found using standard curves that have been derived by fitting against the psychoacoustic model. Plots of the parameters used for varying threshold elevations are shown in FIG. 8 and FIG. 9. Here we can see that the desired compression ratio has an “upside down U” shape for varying threshold elevations.

[0651] One embodiment (Shown in FIG. 7) of the present invention comprises a headset 01, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to the signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
G=g0+(P−T)*(1/R−1),
where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behavior, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module, the headset 01 comprising:

• • [0652] a) an audio receiver 13, coupled to the microprocessor, capable of receiving a first audio signal;
  • [0653] b) a speaker 10, coupled to the DSP chip 03;
  • [0654] c) a microphone 11, capable of receiving a second audio signal, which is noise, the microphone disposed upon the headset in a location removed from the locations of the audio receiver 13 and speaker 10, said microphone suitable for minimizing the effect of the speaker's voice on the ambient noise measurement;
  • [0655] d) a digital signal processing (DSP) chip, coupled to the microphone and the speaker, configured to implement a noise correction algorithm 03, the noise correction algorithm comprising:
    • [0656] i. a first Analog to Digital converter 404,
    • [0657] ii. a second Analog to Digital converter 405,
    • [0658] iii. a first Frequency Analysis Module 406,
    • [0659] iv. a second Frequency Analysis Module 407,
    • [0660] v. a first Power Estimator Block 408,
    • [0661] vi. a second Power Estimator Block 409,
    • [0662] vii. a first dBFS to dBSPL Converter 410,
    • [0663] viii. a second dBFS to dBSPL Converter 411,
    • [0664] ix. a parameter estimation module 422, which determines parameter settings for a parallel compression module, according to information stored in memory,
    • [0665] x. a parallel compression module 417, parameterized by threshold T, compression ratio R, and makeup gain g0, where the parameters may by dynamically varied, where the parallel compression module is comprised of:
      • [0666] a. a linear compression module, which applies gains to the first audio signal according to the equation G=g0+(P−T)*(1/R−1), where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
      • [0667] b. a summer, which combines the resulting amplified first audio signal component with the original first audio signal component,
    • [0668] xi. a Frequency Synthesis Module 418, and
    • [0669] xii. a Digital to Analog Converter 419,
      wherein the DSP chip 03 receives a first audio signal which is the signal of interest via the audio receiver,
      whereupon the DSP Chip 03, obtains a second audio signal from the background noise microphone 11,
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First D/A Converter 404 and Second D/A 405 thereby converting the signals into digital format,
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 406 and Second Frequency Analysis Module 407 breaking down each respective audio signal into separate arrays of frequency components,
      whereupon, for each frequency component, the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 408 and the Second Power Estimator 409, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for a source signal P_dBFS(f) and a background signal P_0—dBFS(f) in dBFS,
      whereupon (i) the DSP chip 03 completes a calibration offset such that a first dBFS to dBSPL Converter 410 converts the power value of the source signal from dBFS to dBSPL, and (ii) the DSP chip 03 completes a calibration offset such that a second dBFS to dBSPL Converter 411 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f),
      whereupon, for each frequency component, and at each instant, the DSP chip 03, using the power estimates of the frequency components of the first 002 and second 003 audio signals, obtains parameters for the parallel compression 422, depending on the sound intensity of the second audio signal at each instant,
      whereupon the DSP chip 03, applies the parameters to a parallel compression module 417,
      whereupon the parallel compression module 417 applies a linear compression, according to the selected parameters, to each of the frequency components of the first audio signal 002,
      whereupon the parallel compression module 417 sums each frequency component of the first audio signal 002 with the corresponding compressed audio signal,
      whereupon the DSP chip 03 repeats the process of estimating the power of both the signal of interest and the noise, selecting compression settings, and applying parallel compression for each frequency component of the first audio signal 002,
      whereupon the DSP chip 03 reconstitutes the first audio signal using the Frequency Synthesis Module 418 producing a corrected audio signal,
      whereupon the corrected audio signal is fed through the Digital to Analog Converter 419, resulting in analog corrected audio signal x_out(t);
      whereupon the DSP chip 03 outputs the corrected audio signal to the speaker 10.

[0670] In one possible embodiment, the parameter estimation module 422 is a polynomial continuous curve defining the relationship between the ambient noise loudness and each of the parameters, where polynomial regression is used to find parameters which define this curve, where these parameters are then stored in memory, so that during processing the parameters can then be computed using a polynomial equation.

[0671] In one possible embodiment, one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P−T′, where PN is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.

[0672] In some embodiments, the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
G=g0+(dP−T′)/R−dP

[0673] where dP is the difference between the signal loudness and noise loudness in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.

[0674] In the aforementioned embodiment, a fixed set of parameters ca be used. The fixed parameters may fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.

[0675] In some embodiments, the First and Second Frequency Analysis Modules are composed of digital filter banks which are filters that break the signal down into various sub-band channels x(t) and x_—0(t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module is a summer operation that combines the various channels into output xout(t).

[0676] In some embodiments, the First and second Frequency Analysis Modules are digital filter banks composed of wavelet filters.

[0677] In some embodiments, the various sub-band channels correspond to critical bands of hearing.

[0678] In some embodiments, the First and Second Frequency Analysis Modules, are composed of FFT coprocessors that perform Fourier transforms on the input signals producing X(t) and X0(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module 418 is an FFT coprocessor that performs an inverse Fourier transform, resulting in output signal xout(t).

[0679] In some embodiments, the frequency components may be grouped together to approximate sub-bands correspond to critical bands of hearing.

[0680] In some embodiments, the processing chain, from output of channels at the digital filter bank to adding gains to channels via the gain component, is a process that occurs in parallel for all channels of the audio signal, through the use of threaded processes, or through the use of a parallel computing device, such as a multi-core processing unit.

[0681] The present invention features systems for audio correction. The systems may include but are not limited to headsets (and the like) and music players (and the like).

[0682] In some embodiments, the present invention may be multi-channel, such as in a stereo or surround sound system. In this case, gains are computed independently for each channel independently. These gains may be based on one central background noise estimate, or multiple noise estimates at each speaker. Each channel is then amplified by its corresponding gains so that the overall surround effect compensates for the presence of background noise.

[0683] In some embodiments, the present invention features an headset for correcting an audio signal. The apparatus, or audio source as claimed, (e.g., a headset) may, for example, comprise a standard headset receiver, a standard headset microphone, and a standard headset speaker, all of which are well known to one of ordinary skill in the art. The receiver can function to extract an amplitude of a source signal (e.g., an incoming conversation on a headset) at given frequency (or within a frequency range) and the microphone can function to detect an amplitude of a background signal at a corresponding frequency (or within a frequency range).

[0684] In some embodiments, the systems of the present invention can evaluate sounds within pre-determined ranges of frequencies, e.g., any appropriate set or group of ranges. Microphones, and/or receivers and/or the like can collect information for the particular frequency range (the pre-determined frequency range). In some embodiments, a first range is 500 Hz and below, a second range is between 500 Hz and 2 kHz, and a third range is above 2 kHz. In some embodiments a first range is 1 kHz and below and a second range is above 1 kHz. The present invention is not limited to the aforementioned ranges.

[0685] In some embodiments, some elements of the above process flow might be performed using analog circuitry before being input to the microprocessor.

[0686] In some embodiments, Analog-to-digital and digital-to-analog converters are assumed to exist at interface points between analog elements and digital elements of any embodiment.

[0687] In some embodiments, programmable logic devices, including but not limited to FPGAs, may be used to implement parts of the processing shown in FIG. 3, with appropriate interfaces implemented between the microprocessor and the programmable logic devices.

[0688] In some embodiments, referring to the means of extracting amplitudes of source signals or background signals: at the output of each filter which selects a band of adjacent frequencies, an electrical circuit to effect extraction of amplitudes could be the use of a diode, or rectifier bridge, fed by each signal which, in turn, feeds a simple series resistor and a capacitor in shunt to create an integrator of the absolute value of the signal. The integrator would accumulate rectified current for some period of time before being applied to the loudness response curve selection circuit, after which the capacitor would be shorted to discharge the accumulated current, in anticipation of the next cycle of integration. Alternatively, the capacitor could be allowed to discharge continuously through a higher valued resistor in shunt, so that charge can be bled off more slowly than it is accumulated, thereby forming a running exponential average integrator.

[0689] In some embodiments, more sophisticated version of amplitude estimation could send the signals through a squarer circuit ahead of the RC integrator in order to accumulate an estimate of the power in the signal. The resulting integrated current would be an estimate of the RMS power at the end of the integration period. This power estimate could then feed the gain correction curve selection circuit.

[0690] In some embodiments, the output of each integrator can be fed to an analog differencing amplifier (operational amplifier, or OpAmp) to subtract the estimated amplitude of the noise signal from the estimated amplitude of the signal of interest. The output of this difference amplifier can be used to establish a dynamic compressor threshold. The output of the compressor circuit (a voltage controlled amplifier, or VCA) can be added to the dry signal to effect a nonlinear correction. A simple linear compressor (VCA) adjusted with suitable make-up gain and compression ratio can produce, in sum with the dry signal, an approximation of the required nonlinear correction gain.

[0691] The present invention is not limited to the aforementioned examples of means of extracting amplitudes.

[0692] Without wishing to limit the present invention to any theory or mechanism, it is believed that the present invention is advantageous because the present invention features a maximum output volume, for example the source signal will be amplified to only a certain degree. This can help protect against damage to the user's hearing. Furthermore, the systems frequently detect the background noise (e.g., 10-40 times per second, 1,000 times per second, etc.), allowing for the systems to react quickly to changes in the background noise. Furthermore, the systems utilize non-linear corrections, thus, loud signals are not amplified much if at all.

[0693] In some embodiments, the systems of the present invention can allow for microphone sensitivity to be altered. For example, a particular user may wish to have more correction or less correction. For example, the microphone may be in someone's pocket, or else exposed directly to the ambient noise sounds. If the microphone becomes shielded, it will produce a weaker response to the noise than if it were directly exposed. Hence the microphone gain can be increased to compensate for the shielding. Also, each of us has a slightly different response to noise levels, some people being more sensitive to it than others. A variable gain on the microphone can help to compensate for individual differences of loudness perception. Finally microphones exhibit varying characteristics depending on their specific construction. A variable microphone gain in the software can help to normalize the response from different microphones.

[0694] In some embodiments, the systems of the present invention can allow for signals to be corrected prior to transmission or recording, such that ambient noise in the recorded or received signal is retained, but the spectral composition of the primary source signal is modified to compensate, as an alternative to standard noise reduction techniques.

[0695] In some embodiments, the headset of the present invention may be a form fitting insulated helmet, such as may be worn by military personnel.

[0696] In some embodiments the microphone of the present invention may be a directional microphone or an omnidirectional microphone, in varying embodiments.

[0697] In some embodiments, the present invention may incorporate sound sensors and means for identifying classes of sounds in the ambient noise environment (i.e. aircraft noise).

[0698] In some embodiments the present invention may be capable of accepting voice activated commands.

[0699] In some embodiments the present invention may be equipped with interconnection ports for allowing the headset to communicate with external devices.

[0700] In some embodiments the present invention may be equipped with a microphone boom, either for allowing the wearer to place the speaking microphone close to his mouth, or to place the noise sensing microphone away from his mouth.

[0701] In some embodiments, the present invention may include an Intra-concha ear piece, which is configured to be positioned inside the concha—the cavity surrounding the opening to the ear canal, or a supranatual headset in which the ear piece is configured to rest against the pinna, or a circumatural headset in which the ear piece is configured to surround the pinna.

[0702] In some embodiments, steps for converting analog audio input signals to digital input signals can be bypassed where the invention utilizes digital input audio sources capable of receiving digital audio signals and transmitting the same to the processor.

[0703] As used herein, the term “hearing impaired” refers to physical hearing impairment, as in the traditional sense of hearing loss, and/or, can also refer to temporary hearing loss caused by a background noise or other temporary hearing loss factors. One of the novel discoveries addressed by the present invention is that a individual with normal hearing can in fact experience hearing impairment due to background noise, or background signals interfering with source noise or source signals.

[0704] As used herein, the term “about” refers to plus or minus 10% of the referenced number. For example, an embodiment wherein the frequency is about 1,000 Hz includes an embodiment wherein the frequency is 900 to 1,100 Hz.

[0705] Various modifications of the invention, in addition to those described herein, will be apparent to those skilled in the art from the foregoing description. Such modifications are also intended to fall within the scope of the invention. Each reference cited in the present application is incorporated herein by reference in its entirety.

[0706] Although there has been shown and described the preferred embodiment of the present invention, it will be readily apparent to those skilled in the art that modifications may be made thereto which do not exceed the scope of the invention.

These are not Claims, they are Additional Disclosure:

1. A headset 01, effective for applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, the headset 01 comprising:

[0707] a) an audio input 13, coupled to the microprocessor, capable of receiving a first audio signal;

[0708] b) a speaker 10, coupled to a DSP chip 03;

[0709] c) a microphone 11, capable of receiving a second audio signal;

[0710] d) a digital signal processing (“DSP”) chip coupled to the microphone and the speaker, configured to implement a noise correction algorithm, the noise correction algorithm comprising:

• • [0711] i. a first Analog to Digital converter 204,
  • [0712] ii. a second Analog to Digital converter 205,
  • [0713] iii. a first Frequency Analysis Module 206,
  • [0714] iv. a second Frequency Analysis Module 207,
  • [0715] v. a first Power Estimator Block 208,
  • [0716] vi. a second Power Estimator Block 209,
  • [0717] vii. a first dBFS to dBSPL Converter 210,
  • [0718] viii. a second dBFS to dBSPL Converter 211,
  • [0719] ix. a first dBSPL to dB Phon Converter 212,
  • [0720] x. a second dBSPL to dB Phon Converter 213,
  • [0721] xi. a Perceptual model 222,
  • [0722] xii. a Gain Block component 217,
  • [0723] xiii. a Frequency Synthesis Module 218, and
  • [0724] xiv. a Digital to Analog Converter 219,
    wherein the DSP chip 03 receives a first audio signal which is the signal of interest via the audio input 13,
    whereupon the DSP Chip 03, obtains a second audio signal from the background noise microphone 11,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First D/A Converter 204 and the Second D/A 205 thereby converting the signals into digital format,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 206 and Second Frequency Analysis Module 207 breaking down each respective audio signal into separate arrays of frequency components,
    whereupon, for each frequency component, the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the various frequency components, whereby said power values are then fed into the dB Converter to generate power values for a source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS,
    whereupon (i) the DSP chip 03 completes a calibration offset such that a first dBFS to dBSPL Converter 210 converts the power value of the source signal from dBFS to dBSPL, and (ii) the DSP chip 03 completes a calibration offset such that a second dBFS to dBSPL Converter 211 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f),
    whereupon (i) a first dBSPL-dB Phon Converter 212 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 213 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f),
    whereupon the DSP chip 03 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the perceptual model 222 whereupon the DSP chip 03 calculates ΔP(f) in dBPhons,
    whereupon the DSP chip 03 calculates a correction gain of power to the source signal G(f), at the frequency component, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 217 which utilizes the Gain equation,
    whereupon the DSP chip 03 reconstitutes the first audio signal with the corrective gain to each of the source signals corresponding to the frequency components, comprising:

[0725] a) multiplying each of the source signals at the corresponding frequency component according to the corresponding corrective gain ΔP(f),

[0726] b) using the Frequency Synthesis Module 18 to reconstitute the resulting audio signal

whereupon the resulting audio signal is fed through the Digital to Analog Converter 219, resulting in analog signal x_out(t),

whereupon the DSP chip 03 outputs the resulting audio signal to the speaker 10.

[0727] 2. The headset of claim 1, where the First and Second Frequency Analysis Modules 206, 207 are composed of digital filter banks which are filters that break the signal down into various sub-band channels x(t) and x₀(t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module 218 is a summer operation that combines the various channels into output x_out(t).
3. The headset of claim 1, where the First and second Frequency Analysis Modules 206, 207 are digital filter banks composed of wavelet filters.
4. The headset of claim 2, where the various sub-band channels correspond to critical bands of hearing.
5. The headset of claim 1, where the First and Second Frequency Analysis Modules 206,207 are composed of FFT coprocessors that perform Fourier transforms on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module 218 is an FFT coprocessor that performs an inverse Fourier transform, resulting in output signal x_out(t).
6. The headset of claim 5, where the frequency components are grouped together to approximate sub-bands correspond to critical bands of hearing.
7. The headset of claim 1, where the perceptual model 222 consists of a forward model which computes subjectively perceived Sones in terms of Phons, and an inverse model which computes Phons in terms of Sones, allowing the Perceptual model to be combined into the Conductor Equation, the Conductor Equation 316 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,
where this equation may be stated as the desired gain in Phons being the Phons amplification necessary to amplify the sound over background noise so that it's subjectively perceived loudness over background noise is the same as the subjectively perceived loudness without the noise, minus some correcting factor, where the function S(P) is found by solving the Earspring equation under experimental boundary conditions, where the Earspring equation 314,315 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), which is a harmonic oscillator model of the signal processing performed by the ear, the cochlea, and immediate processing by the auditory cortex, where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration for a given frequency channel, identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor, where for a particular frequency of the driving force F(t), which represents a pure sinusoid of intensity P, a solution y(t) to this equation can be found, which can then be converted to <y²>, which represents the perceived loudness S in Sones, resulting in a function S(P),

[0728] $S\left(P\right)=\frac{4\left({\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10},$
where S(P) is the perceived loudness in Sones which the listener experiences for a particular sub-band, and P is the sound intensity impinging on the ear in dBPhons for a particular subband, where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency of the earspring, where the constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary parameters.
8. The headset of claim 7, where the set of boundary parameters used to obtain solutions to the Earspring Equation comprises: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz.
9. The headset of claim 8, where a computational model is used to estimate the Earspring equation solutions S(P) and S⁻¹(S), consisting of quadratic functions such as

[0729] $S\left(P\right)\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring equation and are stored in long term memory.
10. The headset of claim 8, where a computational model is used to estimate ΔP, which consists of a quadratic function

[0730] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Conductor equation 316, for various values of P₀ and are stored and retrieved using a parameter estimation module 323.
11. The headset of claim 10, where the parameter estimation module is a lookup table.
12. The headset of claim 10, where the parameter estimation module is a quadratic function that approximates the optimal parameters across a spectrum of values for the threshold elevation P₀.
13. The headset of claim 8, where a computational model is used to estimate the solutions to the Earspring and/or conductor equations, where the computational model is any numerical function approximator 324 that is fit against the Earspring 314, 315 and Conductor equations 316, as described in claim 6, and approximates their solutions, where these approximators may include, but are not limited to linear regression, neural networks, polynomial regression, logistic regression, kernel regression, factor analysis, spline fitting, piecewise curve fitting, radial basis functions, and any other suitable approximator now known or later discovered.
14. The headset of claim 1, wherein the dBSPL to dBPhon Conversion equation 312,313 is:

[0731] $P_{\mathrm{dBPhon}}\left(F\right)=\frac{P_{\mathrm{dBSPL}}\left(F\right)-\mathrm{ATH}\left(F\right)}{1+\mathrm{kATH}\left(F\right)}$

• • [0732] where
  • [0733] F=frequency
  • [0734] ATH(F)=Absolute Threshold of Hearing
  • [0735] K=− 1/120 when ATH(F)>0
    • [0736] =+ 1/240 when ATH(F)<0
      15. The headset of claim 1, wherein the Gain equation 317 is:
      ΔP_dBSPL(F)=ΔP_dBPhon(F)(1+kATH(F))
  • [0737] where
  • [0738] F=frequency
  • [0739] ATH(F)=Absolute Threshold of Hearing
  • [0740] K=− 1/120 when ATH(F)>0
    • [0741] =+ 1/240 when ATH(F)<0
      16. The headset of claim 1, where the absolute threshold of hearing is computed by the formula:

[0742] $\mathrm{ATH}\left(F\right)=\frac{3.64}{F_{\mathrm{KHz}}^{0.8}}-6.5e^{-0.6{\left(F_{\mathrm{kHz}}-3.3\right)}^2}+0.001F_{\mathrm{kHz}}^4-3.37$
17. The headset of claim 1, where there are a plurality of source audio signals corresponding to the various signals of a stereo or surround sound audio system, and where gains for each channel are independently computed and applied to each channel.
18. The headset of claim 1, where there are a plurality of source audio signals corresponding to the various signals of a stereo or surround sound audio system, and a plurality of background noise signals, one for each channel, and where gains for each channel are independently computed, according to the source and noise signal for each channel, and applied to each channel.
19. The headset of claim 1, where the audio input is a physical audio jack.
20. The headset of claim 1, where the audio input is a RF receiver, such as a Bluetooth receiver.
21. A headset 01, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to the signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
G=g0+(P−T)*(1/R−1),
where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behavior, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module,
the headset 01 comprising:

[0743] a) an audio receiver 13, coupled to the microprocessor, capable of receiving a first audio signal;

[0744] b) a speaker 10, coupled to the DSP chip 03;

[0745] c) a microphone 11, capable of receiving a second audio signal, which is noise, the microphone disposed upon the headset in a location removed from the locations of the audio receiver 13 and speaker 10, said microphone suitable for minimizing the effect of the speaker's voice on the ambient noise measurement;

[0746] d) a digital signal processing (DSP) chip, coupled to the microphone and the speaker, configured to implement a noise correction algorithm 03, the noise correction algorithm comprising:

• • [0747] xiii. a first Analog to Digital converter 404,
  • [0748] xiv. a second Analog to Digital converter 405,
  • [0749] xv. a first Frequency Analysis Module 406,
  • [0750] xvi. a second Frequency Analysis Module 407,
  • [0751] xvii. a first Power Estimator Block 408,
  • [0752] xviii. a second Power Estimator Block 409,
  • [0753] xix. a first dBFS to dBSPL Converter 410,
  • [0754] xx. a second dBFS to dBSPL Converter 411,
  • [0755] xxi. a parameter estimation module 422, which determines parameter settings for a parallel compression module, according to information stored in memory,
  • [0756] xxii. a parallel compression module 417, parameterized by threshold T, compression ratio R, and makeup gain g0, where the parameters may by dynamically varied, where the parallel compression module is comprised of:
    • [0757] a. a linear compression module, which applies gains to the first audio signal according to the equation G=g0+(P−T)*(1/R−1), where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
    • [0758] b. a summer, which combines the resulting amplified first audio signal component with the original first audio signal component,
  • [0759] xxiii. a Frequency Synthesis Module 418, and
  • [0760] xxiv. a Digital to Analog Converter 419,
    wherein the DSP chip 03 receives a first audio signal which is the signal of interest via the audio receiver,
    whereupon the DSP Chip 03, obtains a second audio signal from the background noise microphone 11,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First D/A Converter 404 and Second D/A 405 thereby converting the signals into digital format,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 406 and Second Frequency Analysis Module 407 breaking down each respective audio signal into separate arrays of frequency components,
    whereupon, for each frequency component, the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 408 and the Second Power Estimator 409, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for a source signal P_dBFS(f) and a background signal P_0—dBFS(f) in dBFS,
    whereupon (i) the DSP chip 03 completes a calibration offset such that a first dBFS to dBSPL Converter 410 converts the power value of the source signal from dBFS to dBSPL, and (ii) the DSP chip 03 completes a calibration offset such that a second dBFS to dBSPL Converter 411 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f),
    whereupon, for each frequency component, and at each instant, the DSP chip 03, using the power estimates of the frequency components of the first 002 and second 003 audio signals, obtains parameters for the parallel compression 422, depending on the sound intensity of the second audio signal at each instant,
    whereupon the DSP chip 03, applies the parameters to a parallel compression module 417,
    whereupon the parallel compression module 417 applies a linear compression, according to the selected parameters, to each of the frequency components of the first audio signal 002,
    whereupon the parallel compression module 417 sums each frequency component of the first audio signal 002 with the corresponding compressed audio signal,
    whereupon the DSP chip 03 repeats the process of estimating the power of both the signal of interest and the noise, selecting compression settings, and applying parallel compression for each frequency component of the first audio signal 002,
    whereupon the DSP chip 03 reconstitutes the first audio signal using the Frequency Synthesis Module 418 producing a corrected audio signal,
    whereupon the corrected audio signal is fed through the Digital to Analog Converter 419, resulting in analog corrected audio signal x_out(t);
    whereupon the DSP chip 03 outputs the corrected audio signal to the speaker 10.
    22. The headset of claim 21, where the audio receiver is a physical audio jack.
    23. The headset of claim 21, where the audio receiver is a RF receiver, such as a Bluetooth receiver.
    24. The headset of claim 21, where the First and Second Frequency Analysis Modules 406, 407 are composed of digital filter banks which are filters that break the signal down into various sub-band channels x(t) and x₀(t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module 418 is a summer operation that combines the various channels into output x_out(t).
    25. The headset of claim 21, where the First and second Frequency Analysis Modules 406,407 are digital filter banks composed of wavelet filters.
    26. The headset of claim 22, where the various sub-band channels correspond to critical bands of hearing.
    27. The headset of claim 21, where the First and Second Frequency Analysis Modules 406,407 are composed of FFT coprocessors that perform Fourier transforms on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module 418 is an FFT coprocessor that performs an inverse Fourier transform, resulting in output signal x_out(t).
    28. The headset of claim 21, where the frequency components are grouped together to approximate sub-bands correspond to critical bands of hearing.
    29. The headset of claim 21, where discrete sets of parameters are found by dividing the range of possible noise volumes into discrete levels, and where sets of parameters are found for each discrete noise loudness level, where the parameter estimation module 422 is a lookup table, where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.
    30. The headset of claim 21, where the parameter estimation module 422 is a polynomial continuous curve defining the relationship between the ambient noise loudness and each of the parameters, where polynomial regression is used to find parameters which define this curve, where these parameters are then stored in memory, so that during processing the parameters can then be computed using a polynomial equation.
    31. The headset of claim 21, where one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.
    32. The headset of claim 21, where the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
    G=g0+(dP−T′)/R−dP
    where dP is the difference between the signal loudness and noise loudness in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.
    33. The headset of claim 32, where a fixed set of parameters is used.
    34. The headset of claim 33, where the fixed parameters fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.
    35. The headset of claims 1 and 21, where the headset has two speakers, one for each ear.
    36. The headset of claims 1 and 21, where the headset has a second microphone for the reception of speech of the user.

[0761] The present invention relates to the design of personal communications devices and systems designed to enhance personal communications device signals in the presence of noise.

[0762] The term “personal communication device” is used throughout this document to refer to a device which has the capacity to send and receive was wide variety of audio signals between a remote system and a network infrastructure.

[0763] Mobile phones are becoming more sophisticated, including features such as streaming audio, streaming video and radio reception. These devices are used in a wide variety of environments with rapidly changing noise conditions. It is therefore desirable for the phone to have both ability to dynamically adjust sound output in response to local ambient noise. However, uniformly increasing the volume of a signal often results in some part of a signal being amplified too much or too little, since the human ear is not a simple linear mechanism and ambient noise frequency content is not uniform.

[0764] There are many systems for the application of filtering to noise suppression in an audio signal. In most cases, these inventions relate to the removal of noise that is present in the source signal from the origin or introduced into the signal through processing and transmission. Various forms of filtering may be applied which suppresses the noise signal in whole or in part. Generally, these systems have adverse impacts upon the quality of the original signal. However, these systems do not address noise in the environment of the listener, which cannot be filtered.

[0765] Conversely, systems for the suppression of noise in the listener's environment also exist. These systems generally use noise cancellation to remove the disrupting external signal by adding sound projected through headphones which has the effect of countering the sound waves produced by the noise. In this case, the noise is completely canceled and listener is generally unaware of the existence of the external noise, a result which can reduce the awareness of the listener to potential dangers in the environment.

[0766] In some prior art systems, dynamic volume compensation may be used to raise the volume of a source signal of interest over ambient background noise. However, these systems increase the gain in a spectrally uniform manner, raising the volume of all frequency components equally. This effect can distort the perception of music and speech due to the non-linear behavior of the human ear with respect to frequency and volume.

[0767] Microphones and mechanical systems (e.g., computer software) can measure dBSPL; a sound (e.g., 40 dBSPL) at particular frequency (e.g., 1 kHz) sounds just as loud as the sound (e.g., 40 dBSPL) at a different frequency (e.g., 4 kHz) to a microphone or mechanical system. However, our hearing can be affected by the mechanical construction of our outer ear and/or slow variation in sensitivity across the basilar membrane due to fluid damping of the incident waves in the cochlear fluid. The variable sensitivity of human hearing is reflected in the Fletcher-Munson equal loudness contours and the equal-loudness contours from ISO 226:3003 revision (Phons). The equations of the systems of the present invention utilize conversions from dBSPL to Phons and from Phons to dBSPL (incoming sounds levels are converted from dBSPL to Phons for use in the equations, then subsequently the Phons are converted to dBSPL for expression to speakers and headphones. Conversion from dBSPL to Phons and from Phons to dBSPL is in accordance with the Fletcher-Munson equal-loudness contours and the equal-loudness contours from ISO 226:3003 revision.

[0768] Since the human ear dynamically adjusts to sound intensity levels, the presence of background noise alters the threshold at which sounds begin to be perceived. As a result, ambient noise at a given frequency may make sounds at those frequencies that would otherwise be perceptible imperceptible in the presence of ambient noise. In order for the sound to be heard it must be amplified over the background noise. The volume of the ambient noise therefore represents a degree of hearing impairment or baseline threshold elevation over which the sound must be amplified to be perceived.

[0769] This effect varies according to the spectral composition of the noise, that is, spectral components that are sufficiently far from the spectral composition of the noise will remain perceptible. Consequently, using the total intensity of the background noise to raise the intensity of the source uniformly will overly amplify bands which are not affected, possibly raising the volume to damaging levels. In order to amplify only those components which need compensation, the gains to the source signal must vary by spectral band, according to the spectral composition of the noise.

[0770] Moreover, due to the nonlinear response of the human ear, using the spectral intensity of the background noise at a particular band as the gain for the source at that band will produce excessive amplification. In order to compute the correct gain a nonlinear corrective model must be used to compute an appropriate gain for each frequency. The intensity of the background noise as well as the source signal at a given frequency are inputs to this model, and the output is a desired gain for the source signal at that frequency.

[0771] The perceived loudness of sound can be modeled by a harmonic oscillator with spring constant that varies according to the mean power of vibration. This model is called the Earspring model. The Earspring equation 10,11 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration or S(P), identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor. Thus the resonant frequency of the Earspring varies with the amplitude of the <y²> term. Since <y²> is a function of y, this equation is nonlinear.

[0772] For a particular driving force F(t), which we will consider to be a sinusoid at a given frequency and amplitude, for which the amplitude is P in dBPhons, a solution to the earspring equation y(t) can be found, for a particular set of boundary conditions, which is the steady state response to the forcing function. Transforming into the frequency domain we can obtain the mean power of vibration <y²>=½|Y|², which can be rewritten as S(P).

[0773] A possible set of boundary parameters comprises (i) a Sones ratio between the hearing threshold level and the reference level of 40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz; This equation can be solved for numerically though in practice it is more efficient to use a computational model to estimate the solution.

[0774] We can derive an equation for S(P),

[0775] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10}$
where S(P) is the perceived loudness in Sones which the listener experiences, and P is the sound intensity impinging on the ear in dBPhons. Where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency for a ˜40 dBSPL amplitude driving force. The constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary conditions. Note that this formula is independent of frequency although the sound intensity P of any given signal may vary as a function of frequency.

[0776] The present invention features systems for dynamically adjusting audio signals by applying a gain to the signal in a spectrally varying manner to compensate for a noise source in the environment of the listener, such that the sound is perceived by the listener to be unchanged in loudness and spectral composition. The system obtains a threshold elevation for each frequency component by analyzing the spectral composition of the ambient noise. This threshold elevation is then used by a perceptual model of hearing to determine an appropriate gain adjustment for the corresponding frequency component of the source signal which will make that component perceived by the human ear to be just as loud as if the noise were not present. After applying the gains to the source signal, the resulting signal is output to the speaker. The system allows a listener to hear without distortion, over ambient noise, by applying a gain to the source that varies according to the spectral composition of the noise, rather than cancelling the noise, or applying a uniform volume adjustment to the source. The perceived spectral composition of the source is thus adjusted without the removal of the noise signal. Systems may be incorporated into apparatuses including but not limited to personal communications devices and music players.

[0777] Any feature or combination of features described herein are included within the scope of the present invention provided that the features included in any such combination are not mutually inconsistent as will be apparent from the context, this specification, and the knowledge of one of ordinary skill in the art. Additional advantages and aspects of the present invention are apparent in the following detailed description.

The Following Information Regarding Terms is Non-Limiting and Exemplary in Nature for the Purpose of Understanding the Spirit of the Invention

• • [0778] 1. Amplifier: This component can be any amplifier or related device capable of applying gains to an audio signal.
  • [0779] 2. Audio Codec: Audio Encoder/Decoder converts an audio signal to and from its encoded format either for storing in a file or transmission over a network
  • [0780] 3. A/D Converter: Converts audio signals from analog to digital format. In some embodiments, the term A/D Converter is used interchangeably with a means for converting an audio signal from analog to digital form. Analog-to-digital and digital-to-analog converters can be assumed to exist at interface points between analog elements and digital elements of any embodiment. Said means may be electronic, non-electronic or only partially electronic. Said converters, also commonly referred to as Analog to Digital Converters or ADCS, are well known within the art and would be understood by one skilled in the same. Said converters could include but are not limited to direct-conversion ADCs, successive approximation ADCs, a ramp-compare ADC, Wilkinson ADCs, integrating ADCs, Delta-encoded ADCs, pipeline ADCs, time interleaved ADCs, or ADCs with intermediate FM Stage
  • [0781] 4. Audio Device: As used herein, an audio device is any audio device capable of extracting, storing or receiving an audio signal. In some embodiments, the term audio device is used interchangeably with a “means for extracting and/or identifying an audio signal of interest and/or ambient noise signal.” Said means may be, for example, a microphone, receiver, or storage media containing a signal of interest in an audio file and code for reading the signal from an audio file.
  • [0782] 5. Audiogram: One form of defining a person's hearing loss which plots thresholds of hearing relative to a standardized curve that represents ‘normal’ hearing, in dB(HL). An audiogram can be obtained using a behavioral hearing test called Audiometry. The “Tone” test involves presenting different tones at a specific frequency (pitch) and intensity (loudness) and determining thresholds of hearing for each tone where the thresholds correspond to how loud the tone must be presented in order for the person to perceive the tone sound.
  • [0783] 6. Audio Jack: An input by which an audio signal can be received. i.e. from a microphone.
  • [0784] 7. Background signal: The background signal, as used herein, refers to a portion of the second audio signal or the ambient noise signal at the same frequency component as that of the source signal. The source signal and the background signal are paired.
  • [0785] 8. Calibration equation: Calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed. It is represented by: P_dbSPL(F)=P_dBFS(F)+(P_dBSPL0(F)−P_dBFS0(F)) wherein P_dBSPL0 is nominally a value between 65 and 83, and wherein P_dBFS0 is nominally a value between −20 and −12;
    • [0786] where P=Phons amplitude of a source signal of interest
    • [0787] F=Frequency
    • [0788] dBSPL=decibels according to sound pressure
    • [0789] dBFS=decibels relative to full scale
  • [0790] 9. Bark frequency: The bark scale is a psychoacoustic scale for subjective measurements of loudness. The scale is broken down into critical bands of hearing. Critical bands are known frequency bandwidths that correspond to limitations of the human ear to perceive sounds with frequency differences smaller than the critical bandwidths.
  • [0791] 10. D/A Converter: Converts audio signals from digital to analog format. In some embodiments, the term D/A Converter is used interchangeably with a means for converting an audio signal from digital to analog form. Analog-to-digital and digital-to-analog converters can be assumed to exist at interface points between analog elements and digital elements of any embodiment. A D/A converter, or DAC, is capable of converting a digital, usually binary signal code to an analog signal (current voltage or electric charge). DACs may include but are not limited to pulse width modulators, oversampling or interpolating DACs, binary weighted DACs, R-2R Ladder DACs, Successive approximation or cyclic DACs, thermometer coded DACs, and hybrid DACs.
  • [0792] 11. Digital Signal Processor (“DSP”) Chip: A specialized microprocessor with an architecture for the fast operational needs of digital signal processing.
  • [0793] 12. FFT Co-Processor: A specialized microprocessor designed for the purpose of rapidly computing Fourier transforms using a algorithm for computing Fourier transforms commonly known as the ‘Fast Fourier Transform’ (FFT).
  • [0794] 13. Field Programmable Gate Array (“FPGA”): a reprogrammable electronic chip.
  • [0795] 14. Filter banks: An array of band-pass filters that separate an input signal into multiple components, or as used herein, frequency components, each frequency component carrying a single frequency subband of the original signal.
  • [0796] 15. FIR Filter: A filter with an impulse response (or response to any finite length input) of finite duration, because it settles to zero in finite time. This is in contrast to infinite impulse response (IIR) filters, which have internal feedback and may continue to respond indefinitely (usually decaying). FIR filters can be discrete-time or continuous-time, and digital or analog.
  • [0797] 16. Fourier transforms: A mathematical operation that computes the frequency content of an audio signal, taking a given audio signal as input and outputting magnitudes and phases across the frequency spectrum.
  • [0798] 17. Frequency Analysis Model: This model can use a variety of known methods to divide the audio signal of interest into a plurality of frequency components. Such methods may include, but are not limited to methods such as filter banks, Fourier transforms, wavelet transforms or other signal processing techniques. In some embodiments, the term Frequency Analysis Model is used interchangeably with a “means for dividing an audio signal into a plurality of frequency components. Said means may be, for example, a digital filter bank, analog filter bank, FFT Co-processor, or code for computing a Fourier transform analysis to an audio signal on computer-readable media.
  • [0799] 18. Frequency component and Frequency Component Spectrum: A frequency component, as used herein, identifies a portion of the frequency range of an audio signal. A frequency component can may comprises a particular, individual frequency, frequency channels, or frequency bands. A frequency component spectrum is a plurality of frequency components.
  • [0800] 19. Frequency Synthesis Module: A module which the present invention can use to reconstitute the various frequency components of a particular audio signal of interest. The frequency synthesis Module may perform summation of the various frequency components of the signal of interest, or may perform an inverse Fourier transform, or other transform, after said frequency components have been adjusted, to create a new signal or waveform, depending on whether the Frequency analysis module was a filter bank or a fourier transform, or other transform. “Sound Systems: Design and Optimization: Modern Techniques and Tools for Sound System Design and Alignment”, McCarthy, Bob (2010) (Focal Press) Second Edition. In some embodiments, the term Frequency Synthesis Module is used interchangeably with a “means for reconstituting an audio signal”. Said means may be, for example, a summer block, a frequency synthesis module, an FFT-Co processor, an inverse FFT transform, or code for implementing an inverse FFT transform on computer-readable media.
  • [0801] 20. Gain Block component—This component applies the gains to each corresponding frequency component or band. In some embodiments, the term gain block component is used interchangeably with a “means for applying gains to a audio signal.” Said means may be, for example, an amplifier, or gain block component.
  • [0802] 21. Infinite Impulse Response (“IIR”) Filter: A signal processing filter with an impulse response function that is non-zero over an infinite length of time. May be implemented as either digital or analog IIR filters. IIR filters use fewer computing resources because they use fewer taps.
  • [0803] 22. Input Device: An device configured to enable a user to set a parameter b of a personalized sound adjustment model. In some embodiments, the term input device is used interchangeably with “a means for enabling a user to set a parameter b of a personalized sound adjustment model”. Said means may be, for example, a physical input device such as a user adjustable input, a knob, a slider, or any combination thereof. Said means may alternatively be an electronic means or a combination of physical hardware and electronics such as graphical user interface. Said means may further be a touch screen configured to accept a user's choice of parameter b. Additional means would be recognized and understood by one skilled in the relevant art.
  • [0804] 23. Inverse Fourier Transform: A mathematical operation which computes a time domain signal corresponding to a given frequency spectrum.
  • [0805] 24. kHz frequency to critical band center frequency converter: The critical bands of hearing are natural bands in the audio frequency spectrum in the human ear. These bands are characterized by the masking of tones by other tones in the same critical band. There are several experimentally derived frequency scales that attempt to approximate the critical bands including the Bark scale—a psychoacoustic model proposed by Eberhard Zwicker in 1961, and named after Heinrich Barkhausen who proposed the first subjective measurements of loudness. One example of the equation for converting from kHz to critical band frequency is the conversion to Bark frequency:
    CBR(F_kHz)=26.811(1+1.960/F_kHz)−0.53
  • [0806]  In some embodiments, the term kHz frequency to critical band center frequency converter is used interchangeably with a “means for converting a frequency component or series of frequency components to critical band frequency. Said mean may be, for example, a summer.
  • [0807] 25. Linear slope equation: This equation is part of the Personalized sound adjustment model, and is a novel mathematical equation, and one potential embodiment by which threshold elevations are generated for vTuning. In some embodiments, the term Linear Slope equation is used interchangeably with a “means for estimating threshold elevations specific to a particular person.” Said means may be, for example, the Linear Slope equation, an audiogram, or a pattern library.
  • [0808] 26. Lookup Table: A tabular list of parameters indexed by some given variable, such as signal power. Used by the model to select a parameter or set of parameters depending on the current real-time value of a variable. In some embodiments, the term Lookup Table is used interchangeably with a “means for determining threshold elevations by using a lookup table.” Said means may be, for example, computer-readable media code for retrieving a threshold elevation value from a memory device using a user adjustable parameter b and frequency as indexes.
  • [0809] 27. Microphone: The microphone can be any microphone or the like capable of receiving an audio signal.
  • [0810] 28. Parallel Compression: applying a linear gain to a signal which amplifies softer sounds and subsequently adding this amplified sound back in to the original signal. Also known as New York compression. In some embodiments, the term parallel compression is used interchangeably with a “means for computing and/or applying corrective gains to an audio signal where said gains correspond to threshold elevations, using a combination of parallel compression and known psychoacoustic models” or a “means for using parallel compression to approximate a desired non-linear dynamic range compression curve.” Said means can for example, be any known psychoacoustic model such as the model disclosed herein. Said means may for example, be a model where a compression curve defines the output loudness in decibels versus the input loudness in decibels of a signal of interest, wherein the compression is implemented by applying a gain G to the signal of interest according to the following equation:
    G=g0+(P−T)*(1/R−1)
  • [0811]  where G is the gain in dB to be applied to a frequency component of the signal of interest, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of linear compression, whereupon the linearly compressed signal is added together with the original signal of interest, producing a parallel compressed output signal, where the overall amplitude of the input versus the output signal defines a parallel compression curve, where g0, T, and R are thus parameters that vary the shape of the parallel compression curve, and P is the input loudness of the signal of interest. In some embodiments, the term parallel compression is used interchangeably with a “means for applying parallel compression to a signal of interest.” Said means may be, for example, a parallel compression chip or combining a compressed signal with an original signal. In some embodiments, parallel compression may be a “means for applying linear compression to a signal.” Said means may for example be an amplifier. In some embodiments, the term parallel compression is used interchangeably with a “means for controlling the shape of a parallel compression curve.” Said means may be, for example threshold elevations, makeup gains, and compression ratios. In some embodiments, the term parallel compression is used interchangeably with a “means for finding the parameters which best fit a parallel compression curve against a psychoacoustic model”. Said means may be, for example, linear regression, least squares fit, and function approximation. In some embodiments, the term parallel compression is used interchangeably with a “means for storing parameters”. Said means may be, for example, use of a lookup table, flash memory, hard disk, and long-term memory. In some embodiments, the term parallel compression is used interchangeably with a “means for retrieving the parameters to be applied to a parallel compression module.” Said means may be, for example, a lookup table, polynomials, and a parameter estimation module.
  • [0812] 29. Parameter: The parameter is a single mathematical parameter value that when input to the Personalized Sound adjustment, generates may generate a spectrum of threshold elevations.
  • [0813] 30. Pattern Library: A numerical library composed of points and values which measure threshold elevations. In some embodiments, the term Pattern Library is used interchangeably with a “means for creating a pattern library for storing threshold elevations.” Said means may be, for example, obtaining sample threshold elevation patterns for varying frequency for various typical users (users indexed by b) or computer-readable media code for storing a set of patterns in memory indexed by a user adjustable parameter b and a frequency.
  • [0814] 31. Personalized sound adjustment model: A novel mathematical model which enhances sound through a process of non-linear compression. Bark frequency is one of several types of measures of the natural bandwidth of human hearing referred to as critical bandwidth. Critical bands are defined by the distances in frequency in which masking of tones occurs in human ear, roughly correspond to linear position on the cochlea. Thus, the present invention utilizes an equation that, using a single parameter, chosen by a user, generates a spectrum of threshold elevations that generally fall along a straight line when plotted on models using dBHL against Bark frequency where said threshold elevations can be used to adjust a sound signal for persons with hearing loss or persons who simply wish to enhance the audio signal to their preference. In some embodiments, the term personalized sound adjustment model is used interchangeably with “a means for executing a personalized sound adjustment model”. Said means may be computing the model on a microprocessor, a digital signal processor, or using code for executing the model on computer-readable media. In some embodiments, the term personalized sound adjustment model is used interchangeably with a “means for determining threshold elevations from ambient noise.” Said means may be, for example, dividing an ambient noise signal into frequency components, estimating the ambient noise power of a frequency component, and determining threshold elevations corresponding to a frequency component.
  • [0815] 32. Power Estimation Model/Block: estimating the power value of an audio signal. In some embodiments, the term Power Estimation Block/Model is used interchangeably with a “means for estimating a power value of an audio signal.” Audio power estimation can be conducted using a variety of known techniques which would be understood by one skilled in the art. For example, some use a Minimum Statistics approach which is based on tracking minima of a short term power estimate of the audio signal in bands, over time. If a Fourier transform of the signal is used the power estimate is the square of the magnitude of the frequency component.
  • [0816] 33. Psychoacoustic model: Any appropriate psychoacoustic models can be used by the present invention to compute gains needed to amplify sound to overcome the effects of ambient noise on sound perception or gains needed to account for preferred threshold elevations computed for users' of normal hearing. One example of a known psychoacoustic model is as follows:

[0817] $\frac{\sqrt[\alpha ]{P_{\mathrm{SIG}}^{\alpha }+P_{\mathrm{NOISE}}^{\alpha }-P_{\mathrm{THRQ}}^{\alpha }}}{P_{\mathrm{SIG}}}$

• • [0818] 34. where G is the gain ratio, P_SIG is the signal intensity at a frequency in units of power, P_NOISE is the signal intensity of the background noise, P_THRQ is the absolute threshold of hearing, and α=0.2 is a constant. In some embodiments, the term psychoacoustic model is used interchangeably with a “means for generating or computing corrective gains corresponding to predetermined threshold elevations”. Said means may be, for example, additional known psychoacoustic models, perceptual models, computational models, and models utilizing parallel compression fit against known psychoacoustic models.
  • [0819] 35. RF Transceiver: Radio Frequency Transmitter/Receiver—this device interfaces directly with the antenna and can modulate an audio signal with its radio carrier frequency for transmission as well as demodulate a received radio frequency signal
  • [0820] 36. Source signal: The Source signal, as used herein refers to a portion of the first audio signal or the audio signal of interest at a given frequency component.
  • [0821] 37. Speaker: The speaker can be any speaker or the like capable of receiving and projecting a sound signal adjusted or corrected by the present invention.
  • [0822] 38. Storage media: A computer readable media used for storing data, such as a hard disk, RAM, or flash memory
  • [0823] 39. Summer block: Summation occurs when two or more audio signals are combined together to create a new wave form. Summation is the combining of two audio signals at the same frequency—not to be confused with mixing, which involves the combining of audio signals at different frequencies. Electrical summation occurs inside an electrical circuit. In some embodiments, the term summer block is used interchangeably with a “means for combining an audio signal with another audio signal”. Said means may be, for example, a summer.
  • [0824] 40. Threshold elevation: Threshold elevations correlate to the minimum sound pressures required to perceive sounds at various frequencies.
    Additional Terms:
  • [0825] 23. Conductor Equation (Perceptual Model Component): This is a novel equation which computes the gain necessary to amplify sound sufficiently to make the subjectively perceived sound appear to be as loud, to a hearing impaired subject, as if the subject had no hearing loss. In some embodiments, the term Conductor equation is used interchangeably with a means for determining the gain necessary to make a sound perceived to be as loud as if a threshold elevation were not present.
  • [0826] 24. dBSPL: Decibels sound power level—a unit of measure of sound intensity at the ear
  • [0827] 25. dBSPL to dBPhon Conversion equation: This equation converts sound pressure level to Phons in decibels:

[0828] $P_{\mathrm{dBPhon}}\left(F\right)=\frac{P_{\mathrm{dBSPL}}\left(F\right)-\mathrm{ATH}\left(F\right)}{1+\mathrm{kATH}\left(F\right)}$

• • [0829] where
    • [0830] F=Frequency
    • [0831] ATH=Absolute Threshold of Hearing
    • [0832] P=Phons amplitude of a source signal of interest
    • [0833] dBSPL=decibels according to sound pressure
  • [0834] 26. dBFS to dBSPL Converter: This equation converts the audio signal power from audio power in decibels to audio sound pressure level in decibels
  • [0835] 27. Ear Spring Equation/Crescendo (Psychoacoustic Model): This is a novel equation which computes the apparent subjective loudness for a normal subject, of a sound in Sones as a function of the input sound intensity of the sound in Phons. In some embodiments, the term “Ear Spring Equation/Crescendo” is used interchangeably with a “means for computing the loudness in Sones from the intensity in Phons at a person ear.” Said means may be the Earspring Equation and may also be any known psychoacoustic models.
  • [0836] 28. Gain Equation: This equation converts the gain in Phons to a gain in sound pressure level
    ΔP_dBSPL(F)=ΔP_dBPhon(F)(1+kATH(F))
  • [0837]  where:
    • [0838] F=Frequency
    • [0839] ATH=Absolute Threshold of Hearing
    • [0840] P=Phons amplitude of a source signal of interest
    • [0841] dBSPL=decibels according to sound pressure
    • [0842] K=− 1/120 when ATH(F)>0
      • [0843] =+ 1/240 when ATH(F)<0
  • [0844] 29. Output Device: A device configured to output an audio signal. In some embodiments, the term Output Device is used interchangeably with Audio Device. In some embodiments, the term Output Device is used interchangeably with a “mean for outputting an audio signal.” Said means may be, for example, a speaker, a transmitter, computer-readable media code for transmitting an audio signal or audio file and computer-readable media code for writing an audio signal to an audio file.
  • [0845] 30. Phons: Phons are objective units of sound intensity impinging upon the ear. Sound power may be converted from decibels (dBFS) to Phons using a conversion first to dBSPL then to Phons
  • [0846] 31. Psychoacoustic model (Crescendo): This is a novel equation derived from a complex combination of physics and mathematical equations, which computes gains needed to compensate for ambient noise using two major input values, a source signal of interest, and an ambient noise signal. In some embodiments, the term psychoacoustic model is used interchangeably with a “means for determining the gains to be applied to a signal of interest.” Said means may be, for example, a known psychoacoustic model or Crescendo.
  • [0847] 32. Sones: Sones are subjective units of perceived sound loudness.
  • [0848] 33. Square Summer Equation: This equation squares the audio signal and adds the signal up N times, whereby the sum of the signal at the end block, divided by N, generates a mean power value for the audio signal. In some embodiments, this term is used interchangeably with a “means for extracting power values of an ambient noise signal.”

[0849] FIG. 1. Is a drawing of a possible systems architecture for a personal communications device utilizing a DSP chip to implement an ambient noise correction compression algorithm.

[0850] FIG. 2. Is a drawing of a possible systems architecture for a personal communications device utilizing a software application to implement an ambient noise correction compression algorithm.

[0851] FIG. 3. is a drawing of a possible embodiment for the systems architecture of the present invention.

[0852] FIG. 4. Diagrams an implementation of the Perceptual Model of the present invention using analytical solutions to the Earspring model.

[0853] FIG. 5. Diagrams an implementation of the Perceptual Model in which a Numerical approximation to the solutions is made using a function approximator for which parameters have been stored in memory.

[0854] FIG. 6 is a drawing illustrating an application of parallel compression to approximate a psychoacoustic non-linear compressor, where the parameters of the parallel compression are dynamically adjusted.

[0855] FIG. 7 is a drawing of an exemplary embodiment of the present invention, using a dynamically adjusted parallel compression to approximate the desired non-linear compression of the psychoacoustic model.

[0856] FIG. 8 is a plot of the compression ratio needed to make the parallel compression module approximate the psychoacoustic model, as found by fitting the parallel compression curves against the desired model curves.

[0857] FIG. 9 is a plot of the makeup gain needed to make the parallel compression module approximate the psychoacoustic model, as found by fitting the parallel compression curves against the desired model curves.

[0858] FIG. 10 is a drawing of a possible process flow for the present invention, using a computational model that approximated Earspring and conductor equations.

[0859] FIG. 11 is a drawing of a possible process flow for the present invention, using analytical earpsring and conductor equations.

[0860] FIG. 12 is a drawing of a possible process flow for the present invention using parallel compression to approximate the earspring and conductor equation gains to the signal.

[0861] FIG. 1 diagrams a systems architecture for a personal communications device, in some embodiments, with the noise correction system implemented on a DSP processor in an integrated microprocessor core. The noise correction module may take several forms as shown in FIG. 3 and FIG. 7. The personal communications device comprises the following modules:

• • [0862] a) a microprocessor 02, the microprocessor configured to execute a noise correction application 03,
  • [0863] b) an RF transmitter/receiver 04, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest
  • [0864] c) a digital signal processing chip configured to implement an audio codec 05, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by the first and second microphones,
  • [0865] d) a display screen 06, coupled to the microprocessor, upon which the microprocessor may display information for the user to view,
  • [0866] e) a keypad 07, couple to the microprocessor, which the user actuates to enter data into the microprocessor,
  • [0867] f) a data storage/program memory unit 08, coupled to the microprocessor, which the microprocessor uses to store program code and variables,
  • [0868] g) a speaker 10, coupled to the audio codec 05, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device,
  • [0869] h) a microphone 11, capable of receiving a second audio signal, which is noise, disposed upon the device in a location removed from the locations of the first microphone 08, and speaker 09 suitable to minimizing the effect of the speaker's voice on the ambient noise measurement,

[0870] FIG. 2. diagrams a systems architecture for a personal communications device, in some embodiments, with the noise correction system implemented on a external DSP processor on the output signal line. The personal communications device 101 comprising:

• • [0871] a) a microprocessor 102,
  • [0872] b) an RF transmitter/receiver 104, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest
  • [0873] c) a digital signal processing chip configured to implement an audio codec 105, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by the first and second microphones,
  • [0874] d) a display screen 106, coupled to the microprocessor, upon which the microprocessor may display information for the user to view,
  • [0875] e) a keypad 107, couple to the microprocessor, which the user actuates to enter data into the microprocessor,
  • [0876] f) a data storage/program memory unit 108, coupled to the microprocessor, which the microprocessor uses to store program code and variables,
  • [0877] g) a speaker 110, coupled to the audio codec 104, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device,
  • [0878] h) a microphone 111, capable of receiving a second audio signal, which is noise, disposed upon the phone in a location removed from the locations of the first microphone 108, and speaker 109 suitable to minimizing the effect of the speaker's voice on the ambient noise measurement,
  • [0879] i) a second digital signal processing (DSP) chip, coupled to the audio codec, the microphone and the speaker, configured to implement a noise correction algorithm 103

[0880] Using the architecture of FIG. 1 and the noise correction module shown in FIG. 3, in some embodiments, the present invention comprises a personal communications device capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, the personal communications device 01 comprising:

• • [0881] a) a microprocessor 02, the microprocessor configured to execute a noise correction application 03,
  • [0882] b) an RF transmitter/receiver 04, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest
  • [0883] c) a digital signal processing chip configured to implement an audio codec 05, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by the first and second microphones,
  • [0884] d) a display screen 06, coupled to the microprocessor, upon which the microprocessor may display information for the user to view,
  • [0885] e) a keypad 07, couple to the microprocessor, which the user actuates to enter data into the microprocessor,
  • [0886] f) a data storage/program memory unit 08, coupled to the microprocessor, which the microprocessor uses to store program code and variables,
  • [0887] g) a speaker 10, coupled to the audio codec 05, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device,
  • [0888] h) a microphone 11, capable of receiving a second audio signal, which is noise, disposed upon the device in a location removed from the locations of the first microphone 08, and speaker 09 suitable to minimizing the effect of the speaker's voice on the ambient noise measurement,
  • [0889] i) the noise correction algorithm 03, the noise correction algorithm comprising:
    • [0890] i. a first Frequency Analysis Module 206;
    • [0891] ii. a second Frequency Analysis Module 207;
    • [0892] iii. a first Power Estimator Block 208;
    • [0893] iv. a second Power Estimator Block 209;
    • [0894] v. a first dBFS to dBSPL Converter 210;
    • [0895] vi. a second dBFS to dBSPL Converter 211;
    • [0896] vii. a first dBSPL to dB Phon Converter 212;
    • [0897] viii. a second dBSPL to dB Phon Converter 213;
    • [0898] ix. a Perceptual model 222;
    • [0899] x. a Gain Block component 217;
    • [0900] xi. a Frequency Synthesis Module 218;
      whereupon the microprocessor 02 receives a first audio signal which is the signal of interest via the RF transmitter 03, whereupon the microprocessor 01 receives a second audio signal, which is the noise signal via the microphone,
      whereupon the first audio signal 202 and second audio signal 203 are input to the noise correction algorithm,
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 204 and Second Frequency Analysis Module 205 breaking down each respective audio signal into arrays of frequency components;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS 252;
      whereupon (i) the microprocessor 201 completes a calibration offset such that a first dBFS to dBSPL Converter 210 converts the power value of the source signal from dBFS to dBSPL, and (ii) the microprocessor 201 completes a calibration offset such that a second dBFS to dBSPL Converter 211 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f);
      whereupon (i) a first dBSPL-dB Phon Converter 212 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 213 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f);
      whereupon the microprocessor 201 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the perceptual model 22 whereupon the microprocessor 201 calculates ΔP(f) in dBPhons
      whereupon the microprocessor 201 calculates a correction gain of power to the source signal G(f), at the frequency, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 217 which utilizes the Gain equation;
      whereupon the microprocessor 201 reconstitutes the first audio signal with the corrective gain to the source signal corresponding to the frequency, comprising:
  • [0901] a) multiplying the source signal at the frequency according to the corresponding corrective gain ΔP(f);
  • [0902] b) using the Frequency Synthesis Module 218, to reconstitute the resulting audio signal
    whereupon the noise correction algorithm 203 outputs a corrected audio signal to the microprocessor, whereupon the microprocessor sends the encoded audio signal to the audio codec 205,
    whereupon the audio codec decodes the signal into analog form, whereupon the audio codec outputs the signal to the speaker 210.

[0903] Using the architecture of FIG. 2 and the noise correction module shown in FIG. 3, in some embodiments, the present comprises a personal communications device 101, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment,

• • [0904] the personal communications device 101 comprising:
  • [0905] a) a microprocessor 102,
  • [0906] b) an RF transmitter/receiver 104, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest
  • [0907] c) a digital signal processing chip configured to implement an audio codec 105, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by the first and second microphones,
  • [0908] d) a display screen 106, coupled to the microprocessor, upon which the microprocessor may display information for the user to view,
  • [0909] e) a keypad 107, couple to the microprocessor, which the user actuates to enter data into the microprocessor,
  • [0910] f) a data storage/program memory unit 108, coupled to the microprocessor, which the microprocessor uses to store program code and variables,
  • [0911] g) a speaker 110, coupled to the audio codec 104, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device,
  • [0912] h) a microphone 111, capable of receiving a second audio signal, which is noise, disposed upon the device in a location removed from the locations of the first microphone 108, and speaker 109 suitable to minimizing the effect of the speaker's voice on the ambient noise measurement,
  • [0913] i) a second digital signal processing (DSP) chip, coupled to the audio codec, the microphone and the speaker, configured to implement a noise correction algorithm 103, the noise correction algorithm comprising:
    • [0914] i. a first Analog to Digital converter 204;
    • [0915] ii. a second Analog to Digital converter 205;
    • [0916] iii. a first Frequency Analysis Module 206;
    • [0917] iv. a second Frequency Analysis Module 207;
    • [0918] v. a first Power Estimator Block 208;
    • [0919] vi. a second Power Estimator Block 209;
    • [0920] vii. a first dBFS to dBSPL Converter 210;
    • [0921] viii. a second dBFS to dBSPL Converter 211;
    • [0922] ix. a first dBSPL to dB Phon Converter 212;
    • [0923] x. a second dBSPL to dB Phon Converter 213;
    • [0924] xi. a Perceptual model 222;
    • [0925] xii. a Gain Block component 217;
    • [0926] xiii. a Frequency Synthesis Module 218; and
    • [0927] xiv. a Digital to Analog Converter 219;
      whereupon the microprocessor 102 receives a first audio signal which is the signal of interest via the RF transmitter 104,
      whereupon the microprocessor sends the encoded audio signal to the audio codec 105, whereupon the audio codec decodes the signal into analog form,
      whereupon the audio codec outputs the signal to a DSP Chip 103,
      whereupon the DSP Chip 103, obtains a second audio signal from the background noise microphone 111;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First D/A Converter 204 and Second D/A 205 thereby converting the signals into digital format;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 206 and Second Frequency Analysis Module 207 breaking down each respective audio signal into arrays of frequency components;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS;
      whereupon (i) the DSP chip 103 completes a calibration offset such that a first dBFS to dBSPL Converter 210 converts the power value of the source signal from dBFS to dBSPL, and (ii) the DSP chip 103 completes a calibration offset such that a second dBFS to dBSPL Converter 211 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f);
      whereupon (i) a first dBSPL-dB Phon Converter 212 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 213 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f);
      whereupon the DSP chip 103 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the perceptual model 222 whereupon the DSP chip 103 calculates ΔP(f) in dBPhons
      whereupon the DSP chip 103 calculates a correction gain of power to the source signal G(f), at the frequency, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 217 which utilizes the Gain equation;
      whereupon the DSP chip 103 reconstitutes the first audio signal with the corrective gain to the source signal corresponding to the frequency, comprising:
  • [0928] a) multiplying the source signal at the frequency according to the corresponding corrective gain ΔP(f);
  • [0929] b) using the Frequency Synthesis Module 18, to reconstitute the resulting audio signal
    whereupon the corrected audio signal is fed through the Digital to Analog Converter 219, resulting in analog signal x_out(t);
    whereupon the DSP chip 203 outputs the corrected signal to the speaker 210.

[0930] As shown in FIG. 4, in some embodiments, the perceptual model involves analytically computing the value in Sones of the input signal at a given frequency S(P), along with the estimated hearing loss due to background noise in Sones at that frequency S(P₀), and applying these values to the conductor equation to derive a gain in Phons for the signal that will correct for the perceived hearing loss, the Conductor Equation 13 being,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P

[0931] As shown in FIG. 5, in some embodiments, the perceptual model is to use a numerical estimate of these equations, combining the conductor equation and the Earspring equation solutions into one equation. In this embodiment, a function approximator, such as a quadratic estimate, linear regression with quadratic feature, neural network, or other computational model, can be fit against the solutions to the Earspring equation, and the parameters of the model then stored in memory. During operation, these parameters can be retrieved by the microprocessor from a lookup table, and used to compute an approximate value for ΔP for any given P and P_0.

[0932] In some embodiments, as explained by FIG. 10, the present invention can exist in major part on a computer-readable medium 570 storing a set of instructions executable by one or more processors, where the computer-readable medium 570 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of channels which constitute a first audio signal, wherein corrective gains are generated for each channel by solving for a correction value of ΔP for a source signal, wherein said source signal is a channel of the audio signal corresponding to a particular frequency, where ΔP is derived from solving for the output of a perceptual model 522,

[0933] the computer-readable medium comprising:

• • [0934] a. code for reading the first audio signal x(t) 502 from the RF transceiver, said first audio signal comprising a plurality of channels corresponding to varying frequencies,
  • [0935] b. code for obtaining a second audio signal x₀(t) 503 from a microphone, said second audio signal comprising a plurality of channels corresponding to varying frequencies,
  • [0936] c. code for solving for the power P(f) in Phons of the source signal of the source signal, at a frequency:
    • [0937] i. code for extracting a power value of the source signal at a frequency, in dBFS with a first power estimator block 508, comprising:
      • [0938] (A) extracting the first audio signal x(t) 502, said first audio signal comprising a plurality of channels corresponding to varying frequencies f, where said first audio signal is then transmitted to an Analog to Digital Converter 504 thereby converting the signal into digital format;
      • [0939] (B) feeding the first audio signal through a First Frequency Analysis Module 506 where the first audio signal is broken down into separate frequency components;
      • [0940] (C) identifying a particular source frequency component of the first audio signal corresponding to a frequency 550;
      • [0941] (D) generating a power value for the frequency component, whereby said power value is then fed into a dB Converter to generate a power value P_dBFS(f) for the source in dBFS 552;
    • [0942] ii. code for calculating a calibration offset to convert the power value P_dBFS(f) of the source signal in dBFS to a power value P_dBSPL(f) in dBSPL 510 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12;
    • [0943] iii. code for converting the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation 512, producing P(f);
  • [0944] d. code for solving for P₀(f) in Phons of the background signal, at a frequency:
    • [0945] i. code for extracting a power value of the background signal at the frequency, in dBFS with a second Power Estimator Block 509, comprising:
      • [0946] (A) extracting a second audio signal x₀(t) 503, said second audio signal comprising a plurality of channels corresponding to varying frequencies, where said second audio signal is then transmitted to a second Analog to Digital Converter 505 thereby converting the signal into digital format;
      • [0947] (B) feeding the second audio signal through a Second Frequency Analysis Module 507 where the second audio signal is broken down into separate components;
      • [0948] (C) identifying a particular background signal component, corresponding to a particular frequency 550;
      • [0949] (D) obtaining a power value for the background signal component, whereby said power value is then fed into a dB Converter to generate a power value P_0—dBFS(f) for the background signal in dBFS 552;
    • [0950] ii. code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dbSPL(f) in dBSPL using the calibration equation 5 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12;
    • [0951] iii. code for converting the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation 513, producing P₀(f);
  • [0952] e. code for using the computational model 522 to solve for ΔP(f)
  • [0953] f. code for calculating a correction gain of power G(f), for a frequency f, in dBSPL based on the value of ΔP(f) in dB Phons, using a Gain equation 517;
  • [0954] g. code for reconstituting the first audio signal with the corrective gain to the source signal at the frequency, comprising
    • [0955] i. feeding the correction gain to an amplifier wherein the amplifier amplifies the source signal at the frequency by the corrective gain G(f) 553;
    • [0956] ii. combining, using a Frequency Synthesis Module 518, the channels of the corrected signal, reconstituting the corrected signal, producing x_out(t);
  • [0957] h. code for feeding the corrected audio signal through a Digital to Analog Converter 519;
  • [0958] i. code for transmitting the corrected audio signal to an amplifier, to be output through a speaker.

[0959] In some embodiments, as explained by FIG. 11, the present invention can exist in major part on a computer medium, 670 storing a set of instructions executable by one or more processors, where the computer-readable medium 670 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of channels which constitute a first audio signal, wherein corrective gains are generated for each channel by solving for a correction value of ΔP for a source signal, wherein said source signal is a channel of the audio signal corresponding to a particular frequency, where ΔP is derived from a novel Conductor Equation 616,
S(P)=S(P+ΔP)−S(P₀)+S(0) or,
ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,
where S(P)=a Sones equation from an EarSpring equation solution 614, as a function of Phons Pressure, P=a Phons amplitude of a source signal of interest, ΔP=the gain needed to make a source signal appear at proper loudness for impaired hearing, and P₀=an elevated threshold in Phons of impaired hearing, the computer-readable medium comprising:

• • [0960] a) code for reading the first audio signal x(t) 502 from the RF transceiver, said first audio signal comprising a plurality of channels corresponding to varying frequencies,
  • [0961] b) code for obtaining a second audio signal x₀(t) 503 from a microphone, said second audio signal comprising a plurality of channels corresponding to varying frequencies,
  • [0962] c) code for solving for S(P) of the Conductor Equation which comprises:
    • [0963] i. code for extracting a power value of the source signal at a frequency, in dBFS with a first power estimator block 608, comprising:
      • [0964] A. extracting the first audio signal x(t), said first audio signal comprising a plurality of channels corresponding to varying frequencies, where said first audio signal is then transmitted to an Analog to Digital Converter thereby converting the signal into digital format;
      • [0965] B. feeding the first audio signal through a first digital filterbank 606 where the first audio signal is broken down into separate channels corresponding to varying Frequencies f;
      • [0966] C. identifying the source signal, wherein the source signal is a particular channel of the first audio signal corresponding to the frequency 650;
      • [0967] D. feeding the source signal through a squared summer equation 651 whereby the source signal is squared and added up N times, whereby the sum of the signal at the end block divided by N generates a mean power value for the source signal, whereby said power value is then fed into a dB Converter to generate a power value for the source in dBFS 652, P_dBFS(f);
    • [0968] ii. code for calculating a calibration offset to convert the power value of the source signal P_dBFS(f) in dBFS to a power value P_dBSPL(f) in dBSPL 610 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12;
    • [0969] iii. code for converting the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation 612, producing P(f);
    • [0970] iv. code for applying the power value of the source signal P(f), nowin dB Phons, to the EarSpring equation 614 in order to obtain the power value of the source signal in Sones, or S(P(f));
  • [0971] d) code for solving for S(P₀) of the Conductor Equation which comprises:
    • [0972] i. code for extracting a power value P₀(f) of the background signal at the frequency, in dBFS with a second Power Estimator Block 609, comprising
      • [0973] A. extracting a second audio signal x₀(t), said second audio signal comprising a plurality of channels corresponding to varying frequencies, where said second audio signal is then transmitted to a second analog to digital Converter 605 thereby converting the signal into digital format;
      • [0974] B. feeding the second audio signal through a second Digital Filter Bank 607 where the second audio signal is broken down into separate channels X₀(f) corresponding to varying frequencies;
      • [0975] C. identifying the background signal, wherein the background signal is a particular channel X₀(f) of the audio signal corresponding to the frequency;
      • [0976] D. feeding the background signal through a squared summer equation 651 whereby the background signal is squared and added up N times, whereby the sum of the signal at the end block divided by N generates a mean power value for the signal, whereby said power value is then fed into a dB Converter to generate a power value for the background signal in dBFS 652, producing P_0—dBFS(f);
    • [0977] ii. code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dBSPL(f) in dBSPL using the calibration equation 611 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12;
    • [0978] iii. code for converting the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation 613, producing P₀(f);
    • [0979] iv. code for applying the power value of the background signal, now in dB Phons, to the EarSpring equation 615 in order to obtain the power value of the background signal in Sones, or S(P₀);
  • [0980] e) code for using S(P(f)) and S(P₀(f)) to solve for ΔP(f) using the Conductor equation 616 which results in the Conductor equation 616 generating a ΔP(f) in dBPhons;
  • [0981] f) code for calculating a correction gain of power G(f), for a frequency f, in dBSPL based on the value of ΔP(f) in dB Phons, using a Gain equation 617;
  • [0982] g) code for reconstituting the first audio signal with the corrective gain to the source signal at the frequency, comprising:
    • [0983] i. feeding the correction gain to an amplifier wherein the amplifier multiplies the source signal at the frequency by the corrective gain ΔP(f) 653;
    • [0984] ii. combining, using a Summer Block 618, the channels of the corrected audio signal to reconstitute the corrected output x_out(t);
  • [0985] h) code for feeding the first audio signal through a Digital to Analog

[0986] Converter 619;

• • [0987] i) code for transmitting the first audio signal, now in analog format, to the Amplifier 620; and
  • [0988] j) code for causing the Amplifier 620 to output the audio signal, now containing the corrective gain at the frequency, to a Speaker 620 where the Speaker 620 outputs the audio signal with the corrective gain at the frequency;
    wherein the EarSpring equation 614 is [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), where t=time, y(t)=amplitude of vibration, F(t)=driving force, identified as the Phons amplitude in the following equations, <y²>=mean power of vibration or S(P), identified as the Sones power in the following equations, β=damping constant, k=spring constant, and γ=coefficient of power dependence of spring constant, and η=the scale factor;
    wherein the set of boundary parameters comprises (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz;
    wherein the dBSPL to dBPhon Conversion equation 608 is:

[0989] $P_{\mathrm{dBPhon}}\left(F\right)=\frac{P_{\mathrm{dBSPL}}\left(F\right)-\mathrm{ATH}\left(F\right)}{1+\mathrm{kATH}\left(F\right)}$
wherein the Gain equation 613 is:
ΔP_dBSPL(F)=ΔP_dBPhon(F)(1+kATH(F))
where

• • [0990] F=frequency
  • [0991] ATH(F)=Absolute Threshold of Hearing
  • [0992] K=− 1/120 when ATH (F)>0
    • [0993] =+ 1/240 when ATH(F)<0

[0994] In some embodiments, all or part of the present invention could be implemented using a parallel processing architecture such that some or all of the set of computations from step 5 to 10 above may be computed simultaneously using multiple processing units.

[0995] In some embodiments, the absolute threshold of hearing may be computed in one example by the formula:

[0996] $\mathrm{ATH}\left(F\right)=\frac{3.64}{F_{\mathrm{KHz}}^{0.8}}-6.5e^{-0.6{\left(F_{\mathrm{kHz}}-3.3\right)}^2}+0.001F_{\mathrm{kHz}}^4-3.37$

[0997] In some embodiments, as shown in FIG. 6. a Fourier transform of the signal is used so that the gain computation is performed for each element in the frequency spectrum instead of by channel.

[0998] In some embodiments, as shown in FIGS. 5, 8, and 10, a computational model can be used in place of explicit computation of the Earspring and Conductor equations, where the computational model is

[0999] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D and E are parameters which have been fit against solutions to the combined Earspring and Conductor Equations, for each value of P₀ in 5 dB increments, and stored in a lookup table, such that for a given P₀, parameters for A, B, C, D, and E are extracted from the table, and then the computational model is used to compute ΔP given P.

[1000] In some embodiments it is also possible to use separate computational models for S(P) and for S⁻¹(S), with separate sets of parameters fit against the solutions for various boundary parameters and values of P_0.

[1001] In some embodiments the processing chain, from output of channels at the digital filter bank to adding gains to channels via the gain component, is a process that occurs in parallel for all channels of the audio signal, through the use of threaded processes, or through the use of a parallel computing device, such as a multi-core processing unit.

[1002] FIG. 6 shows another alternative embodiment in place of the perceptual model. In this case, a parallel compression module is fit against the Earspring and Conductor equations and used to estimate their responses. In this embodiment, the optimal threshold (T), makeup gain (M) and compression ratio (R) for a plurality of noise signal powers are pre-computed. These parameters are then stored in memory, or approximated using a function approximator across a range of possible noise powers. At run time, these parameters are then computed and retrieved using a parameter estimation module 311 and applied to a linear compression module 312, which then compresses the source audio signal according to the source signal power, and adds the compressed signal back into the source signal. The resulting parallel compression is a non-linear compression of the source that approximates the effect of the Earspring and Conductor equations.

[1003] FIG. 7 demonstrates the way the parallel compression model is implemented in the noise correction module, using the various sub-components of the frequency, where each component has its power estimated and is compressed according to a parameters selected depending on the noise power. The frequency components are then combined to produce an output signal.

[1004] Using the architecture of FIG. 1, in some embodiments, with the noise correction module of FIG. 7 the present invention may comprises a personal communications device 01, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to the signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
G=g0+(P−T)*(1/R−1),
where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behavior, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module,

• • [1005] the personal communications device 01 comprising:
  • [1006] a) a microprocessor 02, the microprocessor being a multi-core processor containing a digital signal processor 03,
  • [1007] b) an RF transmitter/receiver 04, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest
  • [1008] c) a digital signal processing chip configured to implement an audio codec 05, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by the first and second microphones,
  • [1009] d) a display screen 06, coupled to the microprocessor, upon which the microprocessor may display information for the user to view,
  • [1010] e) a keypad 07, couple to the microprocessor, which the user actuates to enter data into the microprocessor,
  • [1011] f) a data storage/program memory unit 08, coupled to the microprocessor, which the microprocessor uses to store program code and variables,
  • [1012] g) a speaker 10, coupled to the audio codec 05, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device,
  • [1013] h) a microphone 11, capable of receiving a second audio signal, which is noise, disposed upon the device in a location removed from the locations of the first microphone 08, and speaker 09 suitable to minimizing the effect of the speaker's voice on the ambient noise measurement,
  • [1014] i) the digital signal processor 03, configured to implement a noise correction algorithm, the noise correction algorithm comprising:
    • [1015] i. a first frequency analysis module 406, configured to extract a plurality of frequency components from the first audio signal;
    • [1016] ii. a second frequency analysis module 407, configured to extract a plurality of frequency components from the second audio signal;
    • [1017] iii. a first power estimation block 408, configured to calculate the sound intensity in decibels of each frequency component of the first audio signal 002;
    • [1018] iv. a second power estimation block 409, configured to calculate the sound intensity of decibels of each frequency component of the second audio signal 003;
    • [1019] v. a first dBFS to dBSPL Converter 410;
    • [1020] vi. a second dBFS to dBSPL Converter 411; a parameter estimation module 422, which determines parameter settings for a parallel compression module, according to information stored in memory,
    • [1021] vii. the parallel compression module 417, parameterized by threshold T, compression ratio R, and makeup gain g0, where the parameters may by dynamically varied, where the parallel compression module is comprised of:
      • [1022] (a) a linear compression module, which applies gains to the first audio signal according to the equation G=g0+(P−T)*(1/R+1), where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
      • [1023] (b) a summer, which combines the resulting amplified first audio signal component with the original first audio signal component
    • [1024] viii. a Frequency Synthesis Module 418;
      whereupon the microprocessor 02 receives a first audio signal which is the signal of interest via the RF transmitter 04, whereupon the microprocessor 01 receives a second audio signal, which is the noise signal via the microphone,
      whereupon the first audio signal 402 and second audio signal 403 are input to the noise correction algorithm,
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 404 and Second Frequency Analysis Module 405 breaking down each respective audio signal into arrays of frequency components;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 408 and the Second Power Estimator 409, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS;
      whereupon (i) the microprocessor 401 completes a calibration offset such that a first dBFS to dBSPL Converter 410 converts the power value of the source signal from dBFS to dBSPL, and (ii) the microprocessor 401 completes a calibration offset such that a second dBFS to dBSPL Converter 411 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f);
      whereupon, for each frequency component, and at each instant, the microprocessor 401, using the power estimates of the frequency components of the first and second audio signals, obtains parameters for the parallel compression 422, depending on the sound intensity of the second audio signal at each instant,
      whereupon the microprocessor 401, applies the parameters to a parallel compression module 417,
      whereupon the parallel compression module 417 applies a linear compression, according to the selected parameters, to each of the frequency components of the first audio signal 002,
      whereupon the parallel compression module 417 sums each frequency component of the first audio signal 002 with the corresponding compressed audio signal,
      whereupon the microprocessor 401 repeats the process of estimating the power of both the signal of interest and the noise, selecting compression settings, and applying parallel compression for each frequency component of the first audio signal 002,
      whereupon the DSP 401 reconstitutes the first audio signal using the Frequency Synthesis Module 418;
      whereupon the DSP 401 outputs a corrected audio signal to the microprocessor,
      whereupon the microprocessor sends the encoded audio signal to the audio codec 05,
      whereupon the audio codec decodes the signal into analog form, whereupon the audio codec outputs the signal to the speaker 10.

[1025] Using the system architecture for the personal communication device shown in FIG. 2, in some the present invention comprises a personal communications device 01, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to the signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
G=g0+(P−T)*(1/R−1),
where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behavior, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module,

• • [1026] the personal communications device 101 comprising:
  • [1027] a) a microprocessor 102,
  • [1028] b) an RF transmitter/receiver 104, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest
  • [1029] c) a digital signal processing chip configured to implement an audio codec 105, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by the first and second microphones,
  • [1030] d) a display screen 106, coupled to the microprocessor, upon which the microprocessor may display information for the user to view,
  • [1031] e) a keypad 107, couple to the microprocessor, which the user actuates to enter data into the microprocessor,
  • [1032] f) a data storage/program memory unit 108, coupled to the microprocessor, which the microprocessor uses to store program code and variables,
  • [1033] g) a speaker 110, coupled to the audio codec 104, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device,
  • [1034] h) a microphone 111, capable of receiving a second audio signal, which is noise, disposed upon the device in a location removed from the locations of the first microphone 108, and speaker 109 suitable to minimizing the effect of the speaker's voice on the ambient noise measurement,
  • [1035] i) a second digital signal processing (DSP) chip, coupled to the audio codec, the microphone and the speaker, configured to implement a noise correction algorithm 103, the noise correction algorithm comprising:
    • [1036] i. a first Analog to Digital converter 404;
    • [1037] ii. a second Analog to Digital converter 405;
    • [1038] iii. a first frequency analysis module 406, configured to extract a plurality of frequency components from the first audio signal;
    • [1039] iv. a second frequency analysis module 407, configured to extract a plurality of frequency components from the second audio signal;
    • [1040] v. a first power estimation block 408, configured to calculate the sound intensity in decibels of each frequency component of the first audio signal 002;
    • [1041] vi. a second power estimation block 409, configured to calculate the sound intensity of decibels of each frequency component of the second audio signal 003;
    • [1042] vii. a first dBFS to dBSPL Converter 10;
    • [1043] viii. a second dBFS to dBSPL Converter 11; a parameter estimation module 422, which determines parameter settings for a parallel compression module, according to information stored in memory,
    • [1044] ix. the parallel compression module 417, parameterized by threshold T, compression ratio R, and makeup gain g0, where the parameters may by dynamically varied, where the parallel compression module is comprised of:
      • [1045] a. a linear compression module, which applies gains to the first audio signal according to the equation G=g0+(P−T)*(1/R+1), where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
      • [1046] b. a summer, which combines the resulting amplified first audio signal component with the original first audio signal component
    • [1047] x. a Frequency Synthesis Module 418; and
    • [1048] xi. a Digital to Analog Converter 419;
      whereupon the microprocessor 102 receives a first audio signal which is the signal of interest via the RF transmitter 104,
      whereupon the microprocessor sends the encoded audio signal to the audio codec 105, whereupon the audio codec decodes the signal into analog form,
      whereupon the audio codec outputs the signal to a DSP Chip 103,
      whereupon the DSP Chip 103, obtains a second audio signal from the background noise microphone 111;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First D/A Converter 404 and Second D/A 405 thereby converting the signals into digital format;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 406 and Second Frequency Analysis Module 407 breaking down each respective audio signal into arrays of frequency components;
      whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 408 and the Second Power Estimator 409, which compute power values for the various channels, whereby said power values are then fed into the dB Converter to generate power values for the source signal P_dBFS(f) and the background signal P_0—dBFS(f) in dBFS;
      whereupon (i) the DSP 103 completes a calibration offset such that a first dBFS to dBSPL Converter 410 converts the power value of the source signal from dBFS to dBSPL, and (ii) the DSP 103 completes a calibration offset such that a second dBFS to dBSPL Converter 411 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f);
      whereupon, for each frequency component, and at each instant, the DSP 103, using the power estimates of the frequency components of the first and second audio signals, obtains parameters for the parallel compression 422, depending on the sound intensity of the second audio signal at each instant,
      whereupon the DSP 103, applies the parameters to a parallel compression module 417,
      whereupon the parallel compression module 417 applies a linear compression, according to the selected parameters, to each of the frequency components of the first audio signal 002,
      whereupon the parallel compression module 417 sums each frequency component of the first audio signal 002 with the corresponding compressed audio signal,
      whereupon the DSP 103 repeats the process of estimating the power of both the signal of interest and the noise, selecting compression settings, and applying parallel compression for each frequency component of the first audio signal 002,
      whereupon the DSP 103 reconstitutes the first audio signal using the Frequency Synthesis Module 418;
      whereupon the corrected audio signal is fed through the Digital to Analog Converter 419, resulting in analog signal x_out(t);
      whereupon the DSP chip 203 outputs the corrected signal to the speaker 10.

[1049] As shown in FIG. 12, the present invention in some embodiments, comprises the alternative parallel compression method on a computer-readable medium 70 storing a set of instructions executable by one or more processors, where the computer-readable medium 1070 is effective in producing proper hearing corrections across a spectrum of frequencies by applying corrective gains of amplitude to a plurality of frequency components which constitute a first audio signal, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the compression is implemented by applying a gain G to the signal of interest according to the following equation:
G=g0+(P−T)*(1/R−1)

• • [1050] where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
    • [1051] whereupon the linearly compressed signal is added to the first audio signal, where the overall amplitude of the input versus the output signal defines a parallel compression curve, where g0, T, and R are thus parameters that vary the shape of the resulting compression curve, where the psychoacoustic model takes as inputs a source signal loudness and a noise signal loudness, where the model computes the gain in sound loudness at a given frequency component required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness level, are determined beforehand and stored in memory, where a different set of parameters is retrieved at each instant depending on the ambient noise signal intensity at each frequency component, where the dynamically changing noise signal intensity results in a dynamically changing selection of parameters for the parallel compression module,
    • [1052] the computer-readable medium comprising:
      • [1053] a) code for, at each frequency component, extracting a power value of the source signal at a frequency component, in dBFS with a first power estimator block 1008; comprising:
        • [1054] i. extracting the first audio signal x(t) 1002, said first audio signal comprising a plurality of frequency components, where said first audio signal is then transmitted to an Analog to Digital Converter 1004 thereby converting the signal into digital format,
        • [1055] ii. feeding the first audio signal through a First Frequency Analysis Module 1006 where the first audio signal is broken down into separate frequency components,
        • [1056] iii. identifying a particular source signal of the first audio signal corresponding to a frequency component 1050,
        • [1057] iv. generating a power value for the source signal, whereby said power value is then fed into a dB Converter to generate a power value P_dBFS(f) for the source signal in dBFS 1052,
        • [1058] v. code for calculating a calibration offset to convert the power value P_dBFS(f) of the source signal in dBFS to a power value P_dBSPL(f) in dBSPL 1010 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,
      • [1059] b) code for extracting a power value of the background signal at the frequency component, in dBFS with a second Power Estimator Block 1009; comprising:
        • [1060] i. extracting a second audio signal x₀(t) 1003, said second audio signal comprising a plurality of frequency components, where said second audio signal is then transmitted to a second Analog to Digital Converter 1005 thereby converting the signal into digital format,
        • [1061] ii. feeding the second audio signal through a Second Frequency Analysis Module 1007 where the second audio signal is broken down into separate frequency components,
        • [1062] iii. identifying a particular background signal, corresponding to the frequency component 1050,
        • [1063] iv. obtaining a power value for the background signal, whereby said power value is then fed into a dB Converter to generate a power value P_0—dBFS(f) for the background signal in dBFS 1052,
        • [1064] v. code for calculating a calibration offset to convert the power value of the background signal P_0—dBFS(f) in dBFS to a power value P_0—dbSPL(f) in dBSPL using the calibration equation 1011 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL₀−dBFS₀) wherein dBSPL₀ is nominally a value between 65 and 83, and wherein dBFS₀ is nominally a value between −20 and −12,
      • [1065] c) code for determining parameters for the parallel compression module 1022, including compression ratio, makeup gain, and threshold, based on the power value of the background signal, where the parameters have previously been fit against solutions to the Earspring and conductor equations, for a variety of background signal power values, and where the parameters may be retrieved at runtime, given a background signal power;
      • [1066] d) code for applying parallel compression to the source signal 1023, using the stored parameters now retrieved, thereby amplifying the source signal according to the desired compression;
      • [1067] e) code for reconstituting the audio signal with the corrective gain to the source signal at the frequency component, comprising combining, using a Frequency Synthesis Module 1018, the parallel compressed frequency components of the audio signal, producing a corrected audio signal, xout(t)

[1068] In a some embodiments, discrete sets of parameters for the parallel compression are found by dividing the range of possible noise volumes into discrete levels, where sets of parameters are found for each discrete noise loudness level, where the parameter estimation module 422 is a lookup table, where during real-time processing the ambient noise intensity is measured and used to select the corresponding set of parameters.

[1069] In some embodiments, the parameter estimation module 422 is a polynomial continuous curve defining the relationship between the ambient noise loudness and each of the parameters, where polynomial regression is used to find parameters which define this curve, where these parameters are then stored in memory, so that during processing the parameters can then be computed using a polynomial equation.

[1070] In some embodiments, one or more parameters (e.g. T) is defined as a function of the ambient noise intensity (e.g. T=P_N−T′, where P_N is noise loudness), such that the parameters stored (e.g. T′) are the parameters of the function defining the relationship between sound intensity (e.g. P_N) and the parameter (e.g. T) of the parallel compression.

[1071] In some embodiments, the gain equation is expressed as a function of the difference between signal intensity and noise intensity:
G=g0+(dP−T′)/R−dP
where dP is the difference between the signal loudness and noise loudness in decibels, and T′ is the threshold offset from the ambient noise intensity to the threshold T.

[1072] In this embodiment, a fixed set of parameters is used. The fixed parameters may fall in the range of 2 to 6 for compression ratio, 15 dB to 25 dB for makeup gain, and −15 to −25 dB for threshold offset.

[1073] In some embodiments the First and Second Frequency Analysis Modules 406, 407 are composed of digital filter banks which are filters that break the signal down into various sub-band channels x(t) and x₀(t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module 418 is a summer operation that combines the various channels into output xout(t).

[1074] In some embodiments, the First and second Frequency Analysis Modules 406,407 are digital filter banks composed of wavelet filters.

[1075] In some embodiments, the various sub-band channels correspond to critical bands of hearing.

[1076] In some embodiments, the First and Second Frequency Analysis Modules 406,407 are composed of FFT coprocessors that perform Fourier transforms on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module 418 is an FFT coprocessor that performs an inverse Fourier transform, resulting in output signal x_out(t).

[1077] In some embodiments, the frequency components may be grouped together to approximate sub-bands correspond to critical bands of hearing.

[1078] In some embodiments, the present invention may be multi-channel, such as in a stereo or surround sound system. In this case, gains are computed independently for each channel independently. These gains may be based on one central background noise estimate, or multiple noise estimates at each speaker. Each channel is then amplified by its corresponding gains so that the overall surround effect compensates for the presence of background noise.

[1079] In some embodiments, the present invention features a personal communications device for correcting an audio signal. The apparatus, or audio source as claimed, (e.g., a personal communications device) may, for example, comprise a standard personal communications device receiver, a standard personal communications device microphone, and a standard personal communications device speaker, all of which are well known to one of ordinary skill in the art. The receiver can function to extract an amplitude of a source signal (e.g., an incoming conversation on a personal communications device) at given frequency (or within a frequency range) and the microphone can function to detect an amplitude of a background signal at a corresponding frequency (or within a frequency range).

[1080] In some embodiments, the systems of the present invention can evaluate sounds within pre-determined ranges of frequencies, e.g., any appropriate set or group of ranges. Microphones, and/or receivers and/or the like can collect information for the particular frequency range (the pre-determined frequency range). In some embodiments, a first range is 500 Hz and below, a second range is between 500 Hz and 2 kHz, and a third range is above 2 kHz. In some embodiments a first range is 1 kHz and below and a second range is above 1 kHz. The present invention is not limited to the aforementioned ranges.

[1081] In some embodiments, some elements of the above process flow might be performed using analog circuitry before being input to the microprocessor.

[1082] In some embodiments, Analog-to-digital and digital-to-analog converters are assumed to exist at interface points between analog elements and digital elements of any embodiment.

[1083] In some embodiments, programmable logic devices, including but not limited to FPGAs, may be used to implement parts of the processing shown in FIG. 3, with appropriate interfaces implemented between the microprocessor and the programmable logic devices.

[1084] In some embodiments, referring to the means of extracting amplitudes of source signals or background signals: at the output of each filter which selects a band of adjacent frequencies, an electrical circuit to effect extraction of amplitudes could be the use of a diode, or rectifier bridge, fed by each signal which, in turn, feeds a simple series resistor and a capacitor in shunt to create an integrator of the absolute value of the signal. The integrator would accumulate rectified current for some period of time before being applied to the loudness response curve selection circuit, after which the capacitor would be shorted to discharge the accumulated current, in anticipation of the next cycle of integration. Alternatively, the capacitor could be allowed to discharge continuously through a higher valued resistor in shunt, so that charge can be bled off more slowly than it is accumulated, thereby forming a running exponential average integrator.

[1085] In some embodiments, a more sophisticated version of amplitude estimation could send the signals through a squarer circuit ahead of the RC integrator in order to accumulate an estimate of the power in the signal. The resulting integrated current would be an estimate of the RMS power at the end of the integration period. This power estimate could then feed the gain correction curve selection circuit.

[1086] In some embodiments, the output of each integrator can be fed to an analog differencing amplifier (operational amplifier, or OpAmp) to subtract the estimated amplitude of the noise signal from the estimated amplitude of the signal of interest. The output of this difference amplifier can be used to establish a dynamic compressor threshold. The output of the compressor circuit (a voltage controlled amplifier, or VCA) can be added to the dry signal to effect a nonlinear correction. A simple linear compressor (VCA) adjusted with suitable make-up gain and compression ratio can produce, in sum with the dry signal, an approximation of the required nonlinear correction gain.

[1087] The present invention is not limited to the aforementioned examples of means of extracting amplitudes.

[1088] Without wishing to limit the present invention to any theory or mechanism, it is believed that the present invention is advantageous because the present invention features a maximum output volume, for example the source signal will be amplified to only a certain degree. This can help protect against damage to the user's hearing. Furthermore, the systems frequently detect the background noise (e.g., 10-40 times per second, 1,000 times per second, etc.), allowing for the systems to react quickly to changes in the background noise. Furthermore, the systems utilize non-linear corrections, thus, loud signals are not amplified much if at all.

[1089] The systems of the present invention can allow for microphone sensitivity to be altered. For example, a particular user may wish to have more correction or less correction. For example, the microphone may be in someone's pocket, or else exposed directly to the ambient noise sounds. If the microphone becomes shielded, it will produce a weaker response to the noise than if it were directly exposed. Hence the microphone gain can be increased to compensate for the shielding. Also, each of us has a slightly different response to noise levels, some people being more sensitive to it than others. A variable gain on the microphone can help to compensate for individual differences of loudness perception. Finally microphones exhibit varying characteristics depending on their specific construction. A variable microphone gain in the software can help to normalize the response from different microphones.

[1090] In some embodiments, the systems of the present invention can allow for signals to be corrected prior to transmission or recording, such that ambient noise in the recorded or received signal is retained, but the spectral composition of the primary source signal is modified to compensate, as an alternative to standard noise reduction techniques.

[1091] In some embodiments, the present invention further comprises an output transducer. Said transducer may be a substance or device, such as a piezoelectric crystal, microphone, or photoelectric cell. Said transducers are effective to convert input energy of one form into output energy of another. An output transducer may be in electronic communication with the microprocessor or DSP.

[1092] In some embodiments, the present invention further comprises one or more sensors for sensing audio signals. Said sensors may be advantageous because they can be placed at one or more locations on the device, thereby increasing the ability of the device to detect audio signals. Said sensors may be in electronic communication with the microprocessor or DSP.

[1093] In some embodiments, the present invention further comprises a remote server for storing user information such as parameters. In some embodiments, the remote server can process audio information. Said feature may be in communication with the microprocessor or DSP.

[1094] In some embodiments, the present invention further comprises a server provider network. For example, the present invention may communicate with a cell phone service network and may make audio adjustments at various locations throughout the network. Said feature greatly enhances the flexibility of the present invention. Said feature may be in communication with the microprocessor or DSP.

[1095] In some embodiments, the present invention further comprises a combination of vTuning and vDBM. For example, the present invention may dynamically adjust audio signals to account for interfering ambient noise while also making a spectral sound adjustment to the signal of interest to personalize sound according to a user's preference. Said feature may be enabled, for example, where the microprocessor or DSP contains both instructions for vTuning and vDBM.

[1096] In some embodiments, the present invention further comprises removable memory devices for storing hearing preferences and profiles. Said memory devices may be, for example, zip drives, memory sticks, memory cards, or the like. Said feature may be in communication with the microprocessor or DSP.

[1097] In some embodiments, the present invention further comprises a composite profile. Said feature may, for example, be electronically stored on the microprocessor or DSP.

[1098] In some embodiments, the present invention further comprises accessory ports. Said ports are effective for allowing the present invention to communicate with additional devices. Said feature may be in communication with the microprocessor or DSP and may be in predetermined locations on the device in places convenient for contact with external devices and connecting peripherals.

[1099] In some embodiments, the present invention further comprises streaming audio, or audio that is constantly received by and presented to an end-user while being delivered by a streaming provider. Said audio may be received be received by the present invention at the audio device.

[1100] In some embodiments, the present invention further comprises streaming video, or video that is constantly received by and presented to an end-user while being delivered by a streaming provider. Said video may be received be received by the present invention at the audio device.

[1101] In some embodiments, the present invention further comprises radio reception. Radio reception may be via the of, for example, an antennae where said antennae is in electronic communication with the microprocessor or DSP.

[1102] In some embodiments, the present invention further comprises resources configured to prompt the user to provide information such as sound alerts, or messages displayed on a screen. Said feature may be instructions encoded to or programmed on the microprocessor or DSP.

[1103] In some embodiments, the present invention further comprises a communication network, effective for sharing information with external devices and systems. Said network may be wired, or wireless.

[1104] In some embodiments, the present invention further comprises an input/output controller. Said device may be a device that interfaces between an input or output device and a computer or hardware device, such as, for example, the microprocessor or DSP.

[1105] In some embodiments, the present invention further comprises one or more of the following peripherals: a USB port, an infrared port, or a storage device. Said peripherals enable the present invention communicate with one or more external devices. Said peripherals may be located at predetermined locations on the personal communications device and be in electronic communication with the microprocessor or DSP.

[1106] In some embodiments, the present invention further comprises common network protocols, which may be stored on the microprocessor or DSP, for sending and receiving information to and from a mobile phone include AMPS (analog signal), TDMA (Time division multiple access), including GSM (General system for mobile), CDMA (code division multiple access), and GPRS (general packet radio service). There are other standards that comprise a sub or super set of these general classifications. The aforementioned are mentioned to include signal means between network and mobile phone that span analog, circuit switched, analog digital and packet based wireless services. Mobile phones adapted to communication with the protocols are provided with signal processing capability that is increasing in power and flexibility.

[1107] In some embodiments, the present invention further comprises a side tone and side tone adder. Side tone may be the effect of sound that is picked up by the telephone's mouthpiece and in real-time introduced at a low level into the earpiece of the same handset, acting as controlled feedback. Adding the side tone may be via an electric adder. Side tone may also be sound diverted from a telephone microphone to the earpiece so that a speaker hears his own voice at the same level and position as that of the respondent.

[1108] In some embodiments, the present invention can be utilized with a standard television set. As such the television may be in electronic communication with the personal communications device through a wired or wireless connection. Said embodiment may allow a standard television to utilize vTuning or VDBM via the personal communications device.

[1109] In some embodiments, the present invention further comprises a maximum gain filter in electronic communication with the microprocessor or DSP and the amplifier or gain block. Said feature may prevent the gains applied to the audio signal of interest to exceed a certain level of gain.

[1110] In some embodiments, the present invention further comprises a peak detector, which may be time varying. Said peak detector may be a series connection of a diode and a capacitor outputting a DC voltage equal to the peak value of an applied AC signal, Said peak detector may be in electronic communication with the microprocessor or DSP, gain block, amplifier, or audio device or any combination thereof.

[1111] In some embodiments, the present invention further comprises a touch screen controller in electronic communication with the microprocessor or DSP.

[1112] In some embodiments, the present invention further comprises an LCD (Liquid Crystal Display). In some embodiments, the present invention further comprises an LCD controller. Said features are effective for display of information pertaining to the present invention to the user and may be in electronic communication with the microprocessor or DSP.

[1113] In some embodiments, the present invention further comprises a baseboard processor. Said feature may be in electronic communication with the microprocessor or DSP, or be utilized in place of the microprocessor or DSP.

[1114] In some embodiments, the present invention further comprises a GPS antenna effective for receiving GPS signals.

[1115] In some embodiments, the present invention further comprises an application processor which may be an integrated Circuit chip similar to the microprocessor or DSP and may control the applications and flow of data on the present invention. Said applications processor may be in communication with and in addition to the microprocessor or DSP, or may exist in place of the microprocessor or DSP.

[1116] In some embodiments, the present invention further comprise a Class D Amp which may be an electronic amplifier where all power devices are operated as binary switches. Said amplifier may be in electronic communication with the microprocessor or DSP.

[1117] In some embodiments, the present invention further comprises a low power audio codec, which may contain vDBM. Said feature may be a computer program implementing an algorithm and said codec may reside or be in electronic communication with the microprocessor or DSP.

[1118] In some embodiments, the present invention further comprises echo cancellation which describes the process by which echo is removed from a voice communication in order to improve sound quality during, for example, a phone call. Said process may be conducted via the DSP or in a software implementation and may use echo suppressors, echo cancelers, or any combination thereof.

[1119] In some embodiments, the present invention further comprises a packet switching network which optimizes data transmission to and from the present invention by grouping data into blocks, called packets, irrespective as to the type of data being transmitted. Said feature may be implemented via a network adapter, transceiver, or the like.

[1120] In some embodiments, steps for converting analog audio input signals to digital input signals can be bypassed where the invention utilizes digital input audio sources capable of receiving digital audio signals and transmitting the same to the processor.

[1121] As used herein, the term “hearing impaired” refers to physical hearing impairment, as in the traditional sense of hearing loss, and/or, can also refer to temporary hearing loss caused by a background noise or other temporary hearing loss factors. One of the novel discoveries addressed by the present invention is that a individual with normal hearing can in fact experience hearing impairment due to background noise, or background signals interfering with source noise or source signals.

[1122] As used herein, the term “about” refers to plus or minus 10% of the referenced number. For example, an embodiment wherein the frequency is about 1,000 Hz includes an embodiment wherein the frequency is 900 to 1,100 Hz.

[1123] Various modifications of the invention, in addition to those described herein, will be apparent to those skilled in the art from the foregoing description. Such modifications are also intended to fall within the scope of the invention. Each reference cited in the present application is incorporated herein by reference in its entirety.

[1124] Although there has been shown and described the preferred embodiment of the present invention, it will be readily apparent to those skilled in the art that modifications may be made thereto which do not exceed the scope of the invention. THESE ARE NOT CLAIMS, THEY ARE ADDITIONAL DISCLOSURE:

[1125] 1. A personal communications device 01, capable of applying corrections to a first audio signal of interest, to compensate for ambient noise in the listener's environment, by applying corrective gains to a plurality of frequency components which constitute the first audio signal, wherein the value of the gains are computed using a nonlinear psychoacoustic model of loudness as a function of sound intensity, as subjectively experienced by a normal human, where this psychoacoustic model takes as inputs the sound intensity of the audio signal P as well as the ambient background noise P₀, wherein corrective gains are generated for each frequency component by solving for a correction gain ΔP for the first audio signal which will suffice to raise the volume of each frequency component of the first audio signal so that the user will perceive it above background noise, by the loudness amount equivalent to how the first audio signal would be perceived in a noiseless environment,
the personal communications device 01 comprising:

[1126] a) a microprocessor 02, the microprocessor being a multicore processor containing a digital signal processor (DSP);

[1127] b) an RF transmitter/receiver 04, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest;

[1128] c) a digital signal processing chip configured to implement an audio codec 05, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received from the RF transmitter/receiver and a microphone;

[1129] d) a display screen 06, coupled to the microprocessor, upon which the microprocessor may display information for the user to view;

[1130] e) an input device 07, couple to the microprocessor, which the user actuates to enter data to the microprocessor

[1131] f) a data storage/program memory unit 08, coupled to the microprocessor, which the microprocessor uses to store program code and variables;

[1132] g) a speaker 10, coupled to the audio codec 05, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device;

[1133] h) the microphone 11, capable of receiving a second audio signal; and

[1134] i) the DSP 03, configured to execute applications for:

• • [1135] i. a first Frequency Analysis Module 206,
  • [1136] ii. a second Frequency Analysis Module 207,
  • [1137] iii. a first Power Estimator Block 208,
  • [1138] iv. a second Power Estimator Block 209,
  • [1139] v. a first dBFS to dBSPL Converter 210,
  • [1140] vi. a second dBFS to dBSPL Converter 211,
  • [1141] vii. a first dBSPL to dB Phon Converter 212,
  • [1142] viii. a second dBSPL to dB Phon Converter 213,
  • [1143] ix. a Perceptual model 222,
  • [1144] x. a Gain Block component 217, and a
  • [1145] xi. a Frequency Synthesis Module 218,
    wherein the microprocessor 02 receives a first audio signal which is the signal of interest via the RF transmitter 03, whereupon the microprocessor 02 receives a second audio signal, which is the noise signal via the microphone,
    whereupon the first audio signal 202 and second audio signal 203 are input to the noise correction algorithm,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 204 and Second Frequency Analysis Module 205 breaking down each respective audio signal into separate arrays of frequency components;
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the various frequency components, whereby said power values are then fed into the dB Converter to generate power values for a source signal P_dBFS(f) and a background signal P_0—dBFS(f) in dBFS, for each frequency component,
    whereupon, for each frequency component, (i) the microprocessor 201 completes a calibration offset such that a first dBFS to dBSPL Converter 210 converts the power value of the source signal of a frequency component from dBFS to dBSPL, and (ii) the microprocessor 201 completes a calibration offset such that a second dBFS to dBSPL Converter 211 converts the power value of the background signal from the frequency component from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f),
    whereupon (i) the first dBSPL-dB Phon Converter 212 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) the second dBSPL-dB Phon Converter 213 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f),
    whereupon the microprocessor 201 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the perceptual model 22 whereupon the microprocessor 201 calculates ΔP(f) in dBPhons,
    whereupon the microprocessor 201 calculates a correction gain of power to the source signal G(f), at the frequency component, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 217 which utilizes the Gain equation,
    whereupon the microprocessor 201 reconstitutes the first audio signal with the corrective gains to the source signals corresponding to each of the frequency components, comprising:

[1146] a) multiplying the source signal at the frequency component according to the corresponding corrective gain ΔP(f), and

[1147] b) using the Frequency Synthesis Module 218, to reconstitute the resulting corrected audio signal

whereupon the noise correction algorithm 203 outputs the corrected audio signal to the microprocessor, whereupon the microprocessor sends the encoded corrected audio signal to the audio codec 205,

whereupon the audio codec decodes the corrected audio signal into analog form,

whereupon the audio codec outputs the corrected audio signal to the speaker 210.

[1148] 2. A personal communications device 01, capable of applying corrections to a first audio signal of interest, to compensate for ambient noise in the listener's environment, by applying corrective gains to a plurality of frequency components which constitute the first audio signal, wherein the value of the gains are computed using a nonlinear psychoacoustic model of loudness as a function of sound intensity, as subjectively experienced by a normal human, where this psychoacoustic model takes as inputs the sound intensity of the audio signal P as well as the ambient background noise P₀, wherein corrective gains are generated for each frequency component by solving for a correction gain ΔP for the first audio signal which will suffice to raise the volume of each frequency component of the first audio signal so that the user will perceive it above background noise, by the loudness amount equivalent to how the first audio signal would be perceived in a noiseless environment,
the personal communications device 101 comprising:

[1149] a) a microprocessor 102;

[1150] b) an RF transmitter/receiver 104, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest;

[1151] c) a digital signal processing chip configured to implement an audio codec 105, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by first and second microphones;

[1152] d) a display screen 106, coupled to the microprocessor, upon which the microprocessor may display information for the user to view;

[1153] e) an input device 107, coupled to the microprocessor, which the user actuates to enter data into the microprocessor;

[1154] f) a data storage/program memory unit 108, coupled to the microprocessor, which the microprocessor uses to store program code and variables;

[1155] g) a speaker 110, coupled to the audio codec 104, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device;

[1156] h) a microphone 111, capable of receiving a second audio signal;

[1157] i) a second digital signal processing (DSP) chip, coupled to the audio codec, the microphone and the speaker, configured to implement a noise correction algorithm 103, the noise correction algorithm comprising:

• • [1158] i. a first Analog to Digital converter 204,
  • [1159] ii. a second Analog to Digital converter 205,
  • [1160] iii. a first Frequency Analysis Module 206,
  • [1161] iv. a second Frequency Analysis Module 207,
  • [1162] v. a first Power Estimator Block 208,
  • [1163] vi. a second Power Estimator Block 209,
  • [1164] vii. a first dBFS to dBSPL Converter 210,
  • [1165] viii. a second dBFS to dBSPL Converter 211,
  • [1166] ix. a first dBSPL to dB Phon Converter 212,
  • [1167] x. a second dBSPL to dB Phon Converter 213,
  • [1168] xi. a Perceptual model 222,
  • [1169] xii. a Gain Block component 217,
  • [1170] xiii. a Frequency Synthesis Module 218, and
  • [1171] xiv. a Digital to Analog Converter 219
    wherein the microprocessor 102 receives a first audio signal which is the signal of interest via the RF transmitter 104,
    whereupon the microprocessor sends the encoded audio signal to the audio codec 105, whereupon the audio codec decodes the signal into analog form, whereupon the audio codec outputs the signal to the DSP Chip 103,
    whereupon the DSP Chip 103, obtains a second audio signal from the background noise microphone 111,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First D/A Converter 204 and Second D/A 205 thereby converting the signals into digital format,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 206 and Second Frequency Analysis Module 207 breaking down each respective audio signal into separate arrays of frequency components,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 208 and the Second Power Estimator 209, which compute power values for the various frequency components, whereby said power values are then fed into the dB Converter to generate power values for a source signal P_dBFS(f) and a background signal P_0—dBFS(f) in dBFS, for each frequency component,
    whereupon (i) the DSP chip 103 completes a calibration offset such that the first dBFS to dBSPL Converter 210 converts the power value of the source signal from dBFS to dBSPL, and (ii) the DSP chip 103 completes a calibration offset such that the second dBFS to dBSPL Converter 211 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f),
    whereupon (i) a first dBSPL-dB Phon Converter 212 converts the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation, giving P(f) and (ii) a second dBSPL-dB Phon Converter 213 converts the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation, giving P₀(f),
    whereupon the DSP chip 103 applies the power value of the source signal and the power value of the background signal, now in dB Phons, to the perceptual model 222 whereupon the DSP chip 103 calculates ΔP(f) in dBPhons,
    whereupon the DSP chip 103 calculates a correction gain of power to the source signal G(f), at the frequency component, in dBSPL, based on the value of ΔP(f) in dB Phons, using the Gain Block component 217 which utilizes the Gain equation,
    whereupon the DSP chip 103 reconstitutes the first audio signal with the corrective gains to each of the source signals corresponding to the frequency components, comprising:

[1172] a) multiplying the source signal at the frequency component according to the corresponding corrective gain ΔP(f),

[1173] b) using the Frequency Synthesis Module 18, to reconstitute the resulting audio signal

whereupon the resulting audio signal is fed through the Digital to Analog Converter 219, generating in analog, signal x_out(t),

whereupon the DSP chip 203 outputs the x_out(t) to the speaker 210.

[1174] 3. A personal communications device 01, capable of applying corrections to a first audio signal of interest, to compensate for ambient noise in the listener's environment, by applying corrective gains to a plurality of frequency components which constitute the first audio signal, wherein the value of the gains are computed using a nonlinear psychoacoustic model of loudness as a function of sound intensity, as subjectively experienced by a normal human, where this psychoacoustic model takes as inputs the sound intensity of the audio signal P as well as the ambient background noise P₀, wherein corrective gains are generated for each frequency component by solving for a correction gain ΔP for the first audio signal which will suffice to raise the volume of each frequency component of the first audio signal so that the user will perceive it above background noise, by the loudness amount equivalent to how the first audio signal would be perceived in a noiseless environment,
the personal communications device 101 comprising:

[1175] a) a microprocessor 102;

[1176] b) an RF transmitter/receiver 104, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest;

[1177] c) a digital signal processing chip configured to implement an audio codec 105, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by first and second microphones;

[1178] d) a display screen 106, coupled to the microprocessor, upon which the microprocessor may display information for the user to view;

[1179] e) an input device 107, coupled to the microprocessor, which the user actuates to enter data into the microprocessor;

[1180] f) a data storage/program memory unit 108, coupled to the microprocessor, which the microprocessor uses to store program code and variables;

[1181] g) a speaker 110, coupled to the audio codec 104, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device;

[1182] h) a microphone 111, capable of receiving a second audio signal;

[1183] i) a computer readable media, the computer-readable medium comprising:

• • [1184] i. code for reading the first audio signal x(t) 502 from an the RF transceiver, said first audio signal comprising a plurality of channels corresponding to varying frequencies,
  • [1185] ii. code for obtaining a second audio signal x0(t) 503 from a microphone 111, said second audio signal comprising a plurality of channels corresponding to varying frequencies,
  • [1186] iii. code for solving for the power P(f) in Phons of the source signal of the source signal, at a frequency:
    • [1187] (1) code for extracting a power value of the source signal at a frequency, in dBFS with a first power estimator block 508, comprising:
      • [1188] (a) extracting the first audio signal x(t) 502, said first audio signal comprising a plurality of channels corresponding to varying frequencies f, where said first audio signal is then transmitted to an Analog to Digital Converter 504 thereby converting the signal into digital format;
      • [1189] (b) feeding the first audio signal through a First Frequency Analysis Module 506 where the first audio signal is broken down into separate frequency components;
      • [1190] (c) identifying a particular source frequency component of the first audio signal corresponding to a frequency 550;
      • [1191] (d) generating a power value for the frequency component, whereby said power value is then fed into a dB Converter to generate a power value PdBFS(f) for the source in dBFS 552;
    • [1192] (2) code for calculating a calibration offset to convert the power value PdBFS(f) of the source signal in dBFS to a power value PdBSPL(f) in dBSPL 510 using a calibration equation wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL0−dBFS0) wherein dBSPL0 is nominally a value between 65 and 83, and wherein dBFS0 is nominally a value between −20 and −12;
    • [1193] (3) code for converting the power value of the source signal to dB Phons by applying a dBSPL to dBPhon Conversion equation 512, producing P(f);
  • [1194] iv. code for solving for P0(f) in Phons of the background signal, at a frequency:
    • [1195] (1) code for extracting a power value of the background signal at the frequency, in dBFS with a second Power Estimator Block 509, comprising:
      • [1196] (a) extracting a second audio signal x0(t) 503, said second audio signal comprising a plurality of channels corresponding to varying frequencies, where said second audio signal is then transmitted to a second Analog to Digital Converter 505 thereby converting the signal into digital format;
      • [1197] (b) feeding the second audio signal through a Second Frequency Analysis Module 507 where the second audio signal is broken down into separate components;
      • [1198] (c) identifying a particular background signal component, corresponding to a particular frequency 550;
      • [1199] (d) obtaining a power value for the background signal component, whereby said power value is then fed into a dB Converter to generate a power value P0_dBFS(f) for the background signal in dBFS 552;
    • [1200] (2) code for calculating a calibration offset to convert the power value of the background signal P0_dBFS(f) in dBFS to a power value P0_dbSPL(f) in dBSPL using the calibration equation 511 wherein calibration comprises configuring a microprocessor to calculate and interpret a baseline relationship between dBFS and dBPSL such that a zero point in dBSPL can be computed and conversion from values in dBFS to values in dBSPL is enabled; wherein the calibration offset equation is dBSPL=dBFS+(dBSPL0−dBFS0) wherein dBSPL0 is nominally a value between 65 and 83, and wherein dBFS0 is nominally a value between −20 and −12;
    • [1201] (3) code for converting the power value of the background signal to dB Phons by applying the dBSPL to dBPhon Conversion equation 513, producing P0(f);
  • [1202] v. code for using the perceptual model 522 to solve for ΔP(f)
  • [1203] vi. code for calculating a correction gain of power G(f), for a frequency f, in dBSPL based on the value of ΔP(f) in dB Phons, using a Gain equation 517;
  • [1204] vii. code for reconstituting the first audio signal with the corrective gain to the source signal at the frequency, comprising
    • [1205] (1) feeding the correction gain to an amplifier wherein the amplifier amplifies the source signal at the frequency by the corrective gain G(f) 553;
    • [1206] (2) combining, using a Frequency Synthesis Module 518, the channels of the corrected signal, reconstituting the corrected signal, producing xout(t);
  • [1207] viii. code for feeding the corrected audio signal through a Digital to Analog Converter 519;
  • [1208] ix. code for transmitting the corrected audio signal to an amplifier 520, to be output through a speaker 521.
    4. The personal communications device of claims 1-3 where the First and Second Frequency Analysis Modules 206, 207 are composed of digital filter banks which are filters that break the signal down into various sub-band channels x(t) and x₀(t), where the power of each channel is found by passing the channel through a Square Summer equation, whereby said signals are squared and added up N times, whereby the sum of the signals at the end block divided by N generates mean power values for the source signal and background signal, and where the Frequency Synthesis Module 218 is a summer operation that combines the various channels into output xout(t).
    5. The personal communications device of claim 1-3, where the First and second Frequency Analysis Modules 206,207 are digital filter banks composed of wavelet filters.
    6. The personal communications device of claim 5 where the various sub-band channels correspond to critical bands of hearing.
    7. The personal communications device of claim 1-3, where the First and Second Frequency Analysis Modules 206,207 are composed of FFT coprocessors that perform Fourier transforms on the input signals producing X(t) and X₀(t), where the power of each component is found by squaring the Fourier component, and the Frequency Synthesis Module 218 is an FFT coprocessor that performs an inverse Fourier transform, resulting in output signal x_out(t).
    8. The personal communications device of claim 7, where the frequency components are grouped together to approximate sub-bands corresponding to critical bands of hearing.
    9. The personal communications device of claim 1-3, where the perceptual model 222 consists of a forward model which computes subjectively perceived Sones in terms of Phons, and an inverse model which computes Phons in terms of Sones, allowing the Perceptual model to be combined into the Conductor Equation, the Conductor Equation 316 being,
    S(P)=S(P+ΔP)−S(P₀)+S(0) or,
    ΔP=S⁻¹(S(P)+S(P₀)−S(0))−P,
    where this equation may be stated as the desired gain in Phons being the Phons amplification necessary to amplify the sound over background noise so that it's subjectively perceived loudness over background noise is the same as the subjectively perceived loudness without the noise, minus some correcting factor, where the function S(P) is found by solving the Earspring equation under experimental boundary conditions, where the Earspring equation 314,315 is written [d²/dt²+2β(d/dt)+k(1+γγ<y²>)]y(t)=ηF(t), which is a harmonic oscillator model of the signal processing performed by the ear, the cochlea, and immediate processing by the auditory cortex, where t=time, y(t)=amplitude of vibration, F(t)=driving force in terms of Phons amplitude or sound intensity, <y²>=mean power of vibration for a given frequency channel, identified as the Sones power in the following equations, β=damping constant, k=spring constant, γ=coefficient of power dependence of spring constant, and η=the scale factor, where for a particular frequency of the driving force F(t), which represents a pure sinusoid of intensity P, a solution y(t) to this equation can be found, which can then be converted to <y²>, which represents the perceived loudness S in Sones, resulting in a function S(P),

[1209] $S\left(P\right)=\frac{\left(4{\hat{\beta }}^2+{\Gamma }_{40}^2\right)}{\left(4{\hat{\beta }}^2+{\left({\Gamma }_{40}S\left(P\right)\right)}^2\right)}{10}^{\left(P-40\right)/10},$
where S(P) is the perceived loudness in Sones which the listener experiences for a particular sub-band, and P is the sound intensity impinging on the ear in dBPhons for a particular subband, where {circumflex over (β)}=β/√{square root over (k)} and where Γ₄₀ represents the mean power of vibration at the resonant frequency of the earspring, where the constants {circumflex over (β)} and Γ₄₀ are found by solving the Earspring equation according to experimental boundary parameters.
10. The personal communications device of claim 9, where the set of boundary parameters used to obtain solutions to the Earspring Equation comprises: (i) a Sones ratio between the hearing threshold level and the reference level of ˜40 dBSPL at 1 kHz, and (ii) the detuning (flattening) of tones by about 75-80 cents as tones range in intensity from ˜40 dBSPL to ˜90 dBSPL near 1 kHz.
11. The personal communications device of claim 1-3, where a computational model is used to estimate Earspring equation solutions S(P) and S⁻¹(S) consisting of quadratic functions such as

[1210] $S\left(P\right)\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Earspring equation and are stored in long term memory.
12. The personal communications device of claim 1-3, where the perceptual model 222 is a computational model 324 used to estimate ΔP, which consists of a quadratic function

[1211] $\Delta P\cong \frac{{\mathrm{AP}}^2+\mathrm{BP}+C}{{\mathrm{DP}}^2+\mathrm{EP}+1},$
where A, B, C, D, and E are parameters which fit the quadratic function against the solution to the Conductor equation 316, as described in claim 6, for various values of P₀ and are stored in a lookup table 323.
13. The personal communications device of claim 1-3, where the perceptual model 222 is any numerical function approximator 224 that is fit against the Earspring 314, 315 and Conductor equations 316, as described in claim 6, and approximates their solutions, where these approximators may include, but are not limited to linear regression, neural networks, polynomial regression, logistic regression, kernel regression, factor analysis, spline fitting, piecewise curve fitting, radial basis functions, and any other suitable approximator now known or later discovered.
14. A personal communications device 01, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to the signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
G=g0+(P−T)*(1/R−1),
where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behavior, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module,
the personal communications device 01 comprising:

[1212] a) a microprocessor 02, the microprocessor being a multi-core processor containing a digital signal processor 03;

[1213] b) an RF transmitter/receiver 04, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest;

[1214] c) a digital signal processing chip configured to implement an audio codec 05, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by first and second microphones;

[1215] d) a display screen 06, coupled to the microprocessor, upon which the microprocessor may display information for the user to view;

[1216] e) an input device 07, coupled to the microprocessor, which the user actuates to enter data into the microprocessor;

[1217] f) a data storage/program memory unit 08, coupled to the microprocessor, which the microprocessor uses to store program code and variables;

[1218] g) a speaker 10, coupled to the audio codec 05, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device;

[1219] h) a microphone 11, capable of receiving a second audio signal;

[1220] i) the digital signal processor 03, configured to implement a noise correction algorithm; the noise correction algorithm comprising:

• • [1221] i. a first frequency analysis module 406, configured to extract a plurality of frequency components from the first audio signal,
  • [1222] ii. a second frequency analysis module 407, configured to extract a plurality of frequency components from the second audio signal,
  • [1223] iii. a first power estimation block 408, configured to calculate the sound intensity in decibels of each frequency component of the first audio signal 002,
  • [1224] iv. a second power estimation block 409, configured to calculate the sound intensity of decibels of each frequency component of the second audio signal 003,
  • [1225] v. a first dBFS to dBSPL Converter 410,
  • [1226] vi. a second dBFS to dBSPL Converter 411,
  • [1227] vii. a parameter estimation module 422, which determines parameter settings for a parallel compression module, according to information stored in memory,
  • [1228] viii. the parallel compression module 417, parameterized by threshold T, compression ratio R, and makeup gain g0, where the parameters may by dynamically varied, where the parallel compression module is comprised of:
    • [1229] (a) a linear compression module, which applies gains to the first audio signal according to the equation G=g0+(P−T)*(1/R−1), where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression,
    • [1230] (b) a summer, which combines the resulting amplified first audio signal component with the original first audio signal component, and
  • [1231] ix. a Frequency Synthesis Module 418
    wherein the microprocessor 02 receives a first audio signal which is the signal of interest via the RF transmitter 04,
    whereupon the microprocessor 01 receives a second audio signal, which is the noise signal via the microphone,
    whereupon the first audio signal 402 and second audio signal 403 are input to the noise correction algorithm,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 404 and Second Frequency Analysis Module 405 breaking down each respective audio signal into separate arrays of frequency components,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 408 and the Second Power Estimator 409, which compute power values for the various frequency components, whereby said power values are then fed into the dB Converter to generate power values for a source signal P_dBFS(f) and a background signal P_0—dBFS(f) in dBFS, for each frequency component,
    whereupon (i) the microprocessor 401 completes a calibration offset such that the first dBFS to dBSPL Converter 410 converts the power value of the source signal from dBFS to dBSPL, and (ii) the microprocessor 401 completes a calibration offset such that the second dBFS to dBSPL Converter 411 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f),
    whereupon, for each frequency component, and at each instant, the microprocessor 401, using the power estimates of the frequency components of the first and second audio signals, obtains parameters for the parallel compression 422, depending on the sound intensity of the second audio signal at each instant,
    whereupon the microprocessor 401, applies the parameters to a parallel compression module 417,
    whereupon the parallel compression module 417 applies a linear compression, according to the selected parameters, to each of the frequency components of the first audio signal 002,
    whereupon the parallel compression module 417 sums each frequency component of the first audio signal 002 with the corresponding compressed audio signal,
    whereupon the microprocessor 401 repeats the process of estimating the power of both the signal of interest and the noise, selecting compression settings, and applying parallel compression for each frequency component of the first audio signal 002,
    whereupon the DSP 401 reconstitutes the first audio signal using the Frequency Synthesis Module 418 resulting in a corrected audio signal,
    whereupon the DSP 401 outputs the corrected audio signal to the microprocessor,
    whereupon the microprocessor sends the encoded corrected audio signal to the audio codec 05,
    whereupon the audio codec decodes the corrected audio signal into analog form,
    whereupon the audio codec outputs the corrected audio signal to the speaker
    15. A personal communications device 01, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to the signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
    G=g0+(P−T)*(1/R−1),
    where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behavior, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module,
    the personal communications device 101 comprising:

[1232] a) a microprocessor 102;

[1233] b) an RF transmitter/receiver 104, coupled to the microprocessor, capable of receiving a first audio signal, which is the signal of interest;

[1234] c) a digital signal processing chip configured to implement an audio codec 105, coupled to the microprocessor, which effectuates the decoding of audio signals for output by a speaker, and the encoding of voice and noise audio received by first and second microphones;

[1235] d) a display screen 106, coupled to the microprocessor, upon which the microprocessor may display information for the user to view;

[1236] e) an input device 107, coupled to the microprocessor, which the user actuates to enter data into the microprocessor;

[1237] f) a data storage/program memory unit 108, coupled to the microprocessor, which the microprocessor uses to store program code and variables;

[1238] g) a speaker 110, coupled to the audio codec 104, disposed upon the device in a location ergonomically suitable for listening through the user's ear when using the device;

[1239] h) a microphone 111, capable of receiving a second audio signal;

[1240] i) a second digital signal processing (DSP) chip, coupled to the audio codec, the RF transmitter/receiver and the speaker, configured to implement a noise correction algorithm 103; the noise correction algorithm comprising:

• • [1241] i. a first Analog to Digital converter 404,
  • [1242] ii. a second Analog to Digital converter 405,
  • [1243] iii. a first frequency analysis module 406, configured to extract a plurality of frequency components from the first audio signal,
  • [1244] iv. a second frequency analysis module 407, configured to extract a plurality of frequency components from the second audio signal,
  • [1245] v. a first power estimation block 408, configured to calculate the sound intensity in decibels of each frequency component of the first audio signal 002,
  • [1246] vi. a second power estimation block 409, configured to calculate the sound intensity of decibels of each frequency component of the second audio signal 003,
  • [1247] vii. a first dBFS to dBSPL Converter 10,
  • [1248] viii. a second dBFS to dBSPL Converter 11,
  • [1249] ix. a parameter estimation module 422, which determines parameter settings for a parallel compression module, according to information stored in memory,
  • [1250] x. the parallel compression module 417, parameterized by threshold T, compression ratio R, and makeup gain g0, where the parameters may by dynamically varied, where the parallel compression module is comprised of:
    • [1251] a. a linear compression module, which applies gains to the first audio signal according to the equation G=g0+(P−T)*(1/R−1), where G is the gain in dB to be applied to a frequency component of the first audio signal, P is the sound intensity of the frequency component in dB, g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio of the linear compression, and
    • [1252] b. a summer, which combines the resulting amplified first audio signal component with the original first audio signal component,
  • [1253] xi. a Frequency Synthesis Module 418, and
  • [1254] xii. a Digital to Analog Converter 419
    wherein the microprocessor 102 receives a first audio signal which is the signal of interest via the RF transmitter 104,
    whereupon the microprocessor sends the encoded audio signal to the audio codec 105, whereupon the audio codec decodes the signal into analog form, whereupon the audio codec outputs the signal to a DSP Chip 103,
    whereupon the DSP Chip 103, obtains a second audio signal from the background noise microphone 111,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First D/A Converter 404 and Second D/A converter 405 thereby converting the signals into digital format,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the First Frequency Analysis Module 406 and Second Frequency Analysis Module 407 breaking down each respective audio signal into separate arrays of frequency components,
    whereupon the first audio signal x(t) and the second audio signal x₀(t) are fed, respectively, through the first Power Estimator 408 and the Second Power Estimator 409, which compute power values for the various frequency components, whereby said power values are then fed into the dB Converter to generate power values for a source signal P_dBFS(f) and a background signal P_0—dBFS(f) in dBFS, at each frequency component,
    whereupon (i) the DSP 103 completes a calibration offset such that the first dBFS to dBSPL Converter 410 converts the power value of the source signal from dBFS to dBSPL, and (ii) the DSP 103 completes a calibration offset such that a second dBFS to dBSPL Converter 411 converts the power value of the background signal from dBFS to dBSPL, producing P_dBSPL(f) and P_0—dBSPL(f),
    whereupon, for each frequency component, and at each instant, the DSP 103, using the power estimates of the frequency components of the first and second audio signals, obtains parameters for the parallel compression 422, depending on the sound intensity of the second audio signal at each instant,
    whereupon the DSP 103, applies the parameters to the parallel compression module 417,
    whereupon the parallel compression module 417 applies a linear compression, according to the selected parameters, to each of the frequency components of the first audio signal 002,
    whereupon the parallel compression module 417 sums each frequency component of the first audio signal 002 with the corresponding compressed audio signal,
    whereupon the DSP 103 repeats the process of estimating the power of both the signal of interest and the noise, selecting compression settings, and applying parallel compression for each frequency component of the first audio signal 002,
    whereupon the DSP 103 reconstitutes the first audio signal using the Frequency Synthesis Module 418,
    whereupon the first audio signal is fed through the Digital to Analog Converter 419, resulting in analog signal x_out(t),
    whereupon the DSP chip 203 outputs the first audio signal to the speaker 10.
    16. A personal communications device 01, capable of applying corrections to incoming audio signals of interest, to compensate for ambient noise in the listener's environment, wherein parallel compression is used to approximate gain curves derived from a psychoacoustic model, wherein the parallel compression is implemented by applying a gain G to the signal of interest according to a linear compression and then combining this compressed signal with the source audio signal, such that the combined parallel compression of the source signal forms a non-linear compression curve, where the linear compression is found by the equation,
    G=g0+(P−T)*(1/R−1),
    where G is the gain in dB to be applied to the signal, where g0 is the makeup gain in dB, T is the threshold in dB, and R is the compression ratio, and where P is the sound intensity in dB of the source audio signal in a channel, such that g0, T, and R are parameters that vary the shape of the resulting parallel compression curve, where the psychoacoustic model takes as inputs the source signal loudness and the noise signal loudness, where the model computes the gain in sound loudness at a given frequency required to correct for the effect of an ambient noise signal, where the parameters g0, T, and R, for a plurality of noise loudness levels, are predetermined by fitting the parallel compression curves against the desired behavior, where a different set of parameters is retrieved at each instant depending on the ambient noise volume in each frequency, where the dynamically changing noise loudness results in a dynamically changing selection of parameters for the parallel compression module,
    the personal communications device 101 comprising: