Audio decoding method and apparatus which recover high frequency component with small computation

[0001] This application is based upon and claims priority from Korean Patent Application No. 2002-40402 filed Jul. 11, 2002, the contents of which are incorporated herein by reference.

[0002] 1. Field of the Invention

[0003] The present invention relates to a method and apparatus which reduce the amount of computation in audio post-processing, and more particularly, to a method and apparatus which recover a high frequency component in MPEG-1 layer 3 audio using stereo mode information detected in a decoder, by selecting a mode depending on a circumstance after dividing stereo modes into two modes. In a first mode, the high frequency component of only one channel is recovered and the recovered high frequency component is used to recover the high frequency component of the other channel when the difference between the two channel signals is small. In a second mode, the high frequency component in only one of every two frames is recovered alternately in the left channel and the right channel and the high frequency component of each of the skipped frames is interpolated based on the high frequency components of the previous frame and the next frame when the difference between the two channel signals is big.

[0004] 2. Description of the Related Art

[0005] In an MPEG-1 layer 3 audio coding method, in order to compress data more efficiently a psychoacoustic model is used in which less bits are allocated to a high frequency component that humans cannot hear. In this case, although the compression ratio improves, high frequency components are lost. Because of the loss of the high frequency components the tone changes and the clarity of sound degrades such that depressed or dull sounds are output. Accordingly, in order to reproduce the tone of the original sound and to improve clarity, a post-processing sound quality improving method for recovering lost high frequency components is needed.

[0006] However, despite their high sound quality conventional post processing sound quality improving methods need an excessive amount of computation to perform the post processing methods. A sub-band replication (SBR) method, which is a widely-known post-processing sound quality improving method, requires about double the number of computation, performed in an MPEG-1 layer 3 decoder.

[0007] A post processing method by which a sound signal decoded in a time domain is provided in order to achieve a high frequency resolution from a frequency spectrum having a combination of a non-high frequency component and noise is disclosed in Korea Patent Laid-Open Gazette No. 2000-75936.

[0008] For a high frequency resolution the method comprises converting a time domain signal that is fast Fourier transform (FFT)-decoded into a frequency domain signal; analyzing an energy distribution in order to detect a disturbance frequency component in the frequency domain signal; detecting the amount of suppression amount on the disturbance frequency component; controlling post filtering; and reverse-converting the data converted by the post filtering.

[0009] However, the method for post processing is performed in a voice decoder to improve perceptual quality of a decoded voice when voice and audio signals are decoded. Although the purposes for improving the resolution of a high frequency component are similar, the steps of the conventional post processing are different from the present invention because the conventional post processing method comprises analyzing a frequency spectrum, and using resolution filtering high frequency. In addition, the effect of reducing the amount of computation in performing the post processing method cannot be achieved in the prior art method.

[0010] To solve the above problems, it is an objective of the present invention to provide a new audio decoding method and apparatus in which, using mode information indicating the similarity between a signal in a left channel and a signal in a right channel in MPEG-1 layer 3 audio data in order to reduce an excessive amount of computation needed in the prior art post processing method, a high frequency component is recovered by selecting a mode depending on a circumstance, after dividing stereo modes into two modes: a first mode in which the high frequency component of only one channel is recovered and the recovered high frequency component is used to recover the high frequency component of the other channel when the difference between the two channel signals is small, and a second mode in which the high frequency component in only one of every two frames is recovered alternately in the left channel and the right channel and the high frequency component of each of the skipped frames is interpolated based on the high frequency components of the previous frame and the next frame when the difference between the two channel signals is big.

[0011] According to an aspect of the present invention, there is provided a high frequency component generation method by which a high frequency component is generated in decoding audio data, the method comprising generating a high frequency component by using mode information, which comprises information on the correlation between a first channel signal and a second channel signal.

[0012] According to another aspect of the present invention, there is provided an audio decoding method which recovers a high frequency component, comprising (a) receiving an encoded audio data, decoding the data, generating an audio signal of a first channel and an audio signal of a second channel, and outputting the audio signals; (b) obtaining from the audio data signal mode information which stores information on the relation between the first channel signal and the second channel signal; (c) if the obtained mode information is a joint stereo mode value indicating that the correlation between the first channel signal and the second channel signal is great, selecting only one of the first channel and the second channel and recovering the high frequency component of the selected channel, and by predetermined modification of the generated high frequency component of the selected channel, recovering the high frequency component of the other channel; and (d) synthesizing the generated high frequency component of the selected channel with the decoded audio signal of the selected channel and outputting the synthesized signal through the selected channel, and synthesizing the generated high frequency component of the other channel with the decoded audio signal of the channel and outputting the synthesized signal through the channel.

[0013] According to another aspect of the present invention, there is provided an audio decoding apparatus which recovers a high frequency component, comprising an audio decoder which receives an encoded audio data, decodes the data, generates an audio signal and outputs the signal; a mode information analyzing unit which analyzes the received encoded audio data and finds a mode information value; a high frequency component generation unit which recovers a high frequency component of each channel according to the mode information value; and an audio synthesizing unit which synthesizes the decoded audio signal with the generated high frequency component and outputs the synthesized signal.

[0021] Referring to FIG. 1, an audio data that is encoded complying with the MPEG-1 layer 3 (MP3) format is input to an MPEG-1 layer 3 decoder 110 which divides the data into a left channel component and a right channel component and generates respective pulse code modulation (PCM) data. The PCM data is input to a high frequency component recovering unit 120 which recovers the high frequency component of each of the left channel and right channel, and outputs recovered PCM audio data.

[0022] Thus, in the prior art post processing method, a high frequency domain is recovered by post processing the left channel signal and the right channel signal, respectively, which are output through the MPEG-1 layer 3 decoder 110.

[0023] However, in most audio signals, there is a high degree of redundancy between a left channel signal and a right channel signal. Accordingly, even in the MPEG-1 layer 3 encoding algorithm, a left channel signal and a right channel signal are not encoded independently in joint stereo mode.

[0024] Therefore, since a left channel signal and a right channel signal are similar to each other and share redundant information, separately post processing the left channel signal and the right channel signal causes unnecessary computation.

[0025] FIG. 2 is a diagram of a post processing apparatus of the present invention.

[0026] An MP3 decoder 210 receives and decodes an MPEG-1 layer 3 audio signal, generates a PCM signal, and outputs the PCM signal. The MP3 audio signal contains mode information. A mode information analyzing unit 220 analyzes mode information in the received MP3 audio signal and sends the value to a high frequency component generation unit 230.

[0027] The high frequency component generation unit 230 generates a high frequency component for each channel according to the mode information value. Depending on whether the mode information has a joint stereo mode value indicating that there is a high degree of similarity between a left channel signal and a right channel signal, or a stereo mode value indicating that there is a low degree of similarity between the two channel signals, the high frequency component generation unit 230 uses different methods to generate a high frequency component. The high frequency component generation process is the same as the prior art high frequency component generation process.

[0028] If the mode value is joint stereo mode (MODE1), it indicates that the left channel signal and the right channel signal are similar. Accordingly, the high frequency component of only one channel is generated and by using the thus generated high frequency component of the channel the high frequency component of the other channel is generated. For example, high frequency components recovered from the left channel signal are multiplied by a correction value with respect to the degree of similarity between two channels and the results are applied as the high frequency components of the right channel signal. Because there is a high degree of similarity between the two signals, this method causes little degradation in sound quality.

[0029] If the mode value is stereo mode (MODE2), it indicates that the difference between the left channel signal and the right channel signal is great. Accordingly, there is no relation between the two channels and the high frequency component of each of the two channels is generated separately. In this case, the high frequency component in only one of every two frames is recovered alternately in the left channel and the right channel and the high frequency component of each of the skipped frames is recovered by interpolation based on the high frequency components of the previous frame and the next frame separately in the left channel and the right channel. MODE2 will be explained later referring to FIG. 5.

[0030] An audio synthesizing unit 240 synthesizes the recovered high frequency components with the decoded PCM signal and outputs the synthesized signal.

[0031] FIG. 3 is a diagram showing the format of an MPEG-1 layer 3 audio stream.

[0032] The MPEG-1 layer 3 audio stream is formed of audio access units (AAUs) 300. The AAU 300 is a minimum unit that can be decoded individually and contains a predetermined number of compressed data samples. The AAU 300 comprises a header 310, an error check field (cyclic redundancy check (CRC)) 320, an audio data field 330 and an auxiliary data field 340.

[0033] The header 310 contains a sync word, ID information, protection bit presence information, bit rate index information, sampling frequency information, padding bit presence information, an individual purpose bit, mode information, mode extension information, copyright information, information indicating original or copy data, and emphasis information.

[0034] The CRC field 320 is optional. The presence of this field is defined in the header 310. The length of this field is 16 bits.

[0035] The audio data field 330 is comprised of compressed audio data.

[0036] Auxiliary data is the data remaining at the end of the audio data when the audio data does not reach the end of one AAU. The auxiliary data field 340 may be comprised of arbitrary data other than MPEG audio data.

[0037] In the post processing method of the present invention, mode information among these information items and information on the difference between a left channel signal and a right channel signal are used to determine the correlation between channels, and according to the determined result, a high frequency component is recovered.

[0038] FIG. 4 is a flowchart showing a post processing method of the present invention.

[0039] In the new post processing method, first, an MPEG-1 layer 3 audio signal is decoded to generate a PCM signal and the generated PCM signal is output (410). Using this output, mode information is obtained (420). It is determined whether the mode information contains a joint stereo mode value indicating there is a high degree of similarity between a left channel signal and a right channel signal or a stereo mode value indicating that there is a low degree of similarity between the two channel signals (430).

[0040] If the result of the determination indicates that the mode value is joint stereo mode (MODE1), it indicates that the left channel signal and the right channel signal are similar. Accordingly, the high frequency component of only one channel is generated (460) and by using the thus-generated high frequency component of the channel the high frequency component of the other channel is generated (470). For example, high frequency components recovered from the left channel signal are used as the high frequency components of the right channel signal without change, or by using information on the difference between the left channel signal and the right channel signal, the high frequency components of the right channel signal are obtained by multiplying the high frequency components by some modification values (470). The modification value, for example, may be a constant. Because the similarity between the two signals is great, this method causes little degradation in sound quality. Finally, the recovered high frequency components are synthesized with the decoded PCM signal and output (480).

[0041] If the result of the determination indicates that the mode value is stereo mode (MODE2), it indicates that the difference between the left channel signal and the right channel signal is great. Accordingly, there is no relation between the two channels and the high frequency component of each of the two channels is generated separately (440). In this case, the high frequency component in only one of every two frames is recovered alternately in the left channel and the right channel and the high frequency component of each of the skipped frames is recovered by interpolation based on the high frequency components of the previous frame and the next frame separately in the left channel and the right channel. Finally, the recovered high frequency components are synthesized with the decoded PCM signal and output (450).

[0042] In general, for most audio signals, the joint stereo mode (MODE1) is applied. That is, since the left channel signal and the right channel signal are similar, this method is applied in many cases such that the amount of computation can be cut in half. Also, when a left channel signal and a right channel signal should be separately encoded because the difference between the two channel signals is great in an audio signal having a high transmission rate of 128 kbps or more, that is, when the two signals fall under MODE2, audio post processing should be applied to the channels separately. However, even in MODE2, the MODE2 method can be used so that the amount of computation can be reduced.

[0043] The new post processing method determines whether a mode is the joint stereo mode indicating that the similarity between the two channel signals is great or the stereo mode indicating that the difference between the two channel signals is great, by using mode information detected in the MPEG-1 layer 3 decoder.

[0044] Thus recovered high frequency components are added to the left channel signal and right channel signal and output.

[0045] FIG. 5 is a diagram showing a method recovering a high frequency component of a right channel.

[0046] First, the high frequency component of the left channel is generated. If the mode information indicates the joint stereo mode, the generated high frequency component of the left channel is used in generating the high frequency component of the right channel; otherwise the high frequency component of the right channel is generated independently of the high frequency component of the left channel.

[0047] FIG. 6 is a diagram showing a method of recovering a high frequency component when the difference between two channel signals is large. That is, FIG. 6 explains the process (440) for separately generating the high frequency components of the left channel and the right channel in FIG. 4. In this case, the high frequency component in only one of every two frames is recovered alternately in the left channel and the right channel and the high frequency component of each of the skipped frames is interpolated based on the high frequency components of the previous frame and the next frame. That is, the high frequency component of the left channel is recovered in the frames of time t1 and the high frequency component of the right channel is recovered in the frames of time t2. The same method is repeatedly performed for times t3, t4, t5, . . . .

[0048] Here, the high frequency component of the left channel and the high frequency component of the right channel in the frames of time t1 are referred to as L_t1 and R_t1, respectively, and in the same manner the high frequency components of the left channel and the right channel in the frames of time t2 are referred to as L_t2 and R_t2, respectively.

[0049] Then, the high frequency component of a skipped frame, that is, the high frequency component of the left channel at time t2, is recovered by interpolation based on the high frequency components of the left channel at time t1 and time t3. That is, L_t2=(L_t1+L_t3)/2. Also, the high frequency component of the right channel at time t3 is recovered by interpolation based on the high frequency components of the right channel at time t2 and time t4. That is, R_t3=(R_t2+R_t4)/2. Finally, the recovered high frequency components are synthesized with the decoded PCM signal and then output.

[0050] The present invention may be embodied in a code, which can be read by a computer, on a computer readable recording medium. The computer readable recording medium includes all kinds of recording apparatuses on which computer readable data are stored.

[0051] The computer readable recording media includes storage media such as magnetic storage media (e.g., ROM's, floppy disks, hard disks, etc.), optically readable media (e.g., CD-ROMs, DVDs, etc.) and carrier waves (e.g., transmissions over the Internet).

[0052] Optimum embodiments have been explained above and are shown. However, the present invention is not limited to the preferred embodiment described above, and it is apparent that variations and modifications by those skilled in the art can be effected within the spirit and scope of the present invention defined in the appended claims. Therefore, the scope of the present invention is not determined by the above description but by the accompanying claims.

[0053] As described above, while it is difficult to apply the prior art post processing method to actual products because of the excessive amount of computation despite the sound quality improvement effect, the present invention can reduce the amount of computation less than half the amount of computation used in the prior art in recovering high frequency components.