Telephone Line Monitoring System

The invention relates generally to a circuit for the characterization of sound. In particular, the invention relates to a circuit to detect voice signals in the presence of supervisory signals on a number of telephone lines.

When a machine-placed telephone call is made to a subscriber, such as is done by a telemarketing computer calling system, it is desirable to have the calling system designed to automatically determine when that call has been answered. This determination must be made within 50 milliseconds so that a waiting operator can be connected to the subscriber by the calling system without any perceivable delay.

Inasmuch as there are typically several seconds between normal ring signals used by the telephone switching network, it is not possible to determine the called party pickup by detecting loss of ring voltage. Furthermore, it is not uncommon to have the called party pickup prior to the initiation of ring signaling as received by the calling system, since ring signals heard by the called party may not be generated by the same signal source that generates ring signals to the calling system. For these reasons, it is apparent that a means of detecting voice is desirable since it may be assumed that the called party will always respond verbally when answering the telephone.

Many of the techniques employed by present telephone line monitoring systems cannot detect or distinguish between voice and telephone tones in a fast and inexpensive manner. For instance, devices that detect and distinguish between voice and tone signals on the line through a reliance on the time delays exhibited by the type of signal are subject to error due to widely ranging variations in tone signals which may occur. Also, devices which employ pattern recognition techniques require a substantial amount of dedicated equipment and, thus, a substantial amount of cost.

U.S. Pat. No. 4,356,348 ("Smith") describes a device for detecting the condition of response of a telephone line by determining the most prevalent time interval between zero voltage crossovers of an input signal, comparing that interval with succeeding time intervals and classifying the input signal as periodic or not. The nature of the input signal is further analyzed to ascertain the condition on the telephone line. However, the mechanism used by the device to detect the response condition requires relatively long sampling periods of the input signal.

U.S. Pat. No. 4,405,833 ("Cave, et al.") describes a device for determining the status of a telephone call by measuring the periods of the low frequency envelope cycles of an incoming signal and evaluating them to determine if the incoming signal comprises call progress tones or voice. Voice is indicated if the periods are not reasonably constant over a standard measurement interval. Analog techniques, such as is utilized by the device, are relatively expensive to implement.

The foregoing problems are obviated by the invention which is a circuit for characterizing a sound signal by sampling the electrical signal produced by an electroacoustic transducer, comprising:

(a) means for detecting the occurrence of each waveform composing a sampled electrical signal;

(b) means for measuring the wavelength of each waveform detected;

(c) means for counting the number of waveforms within each of a plurality of predetermined wavelength groupings; and

(d) means for characterizing the sound signal based on the counts of waveforms within the plurality of predetermined wavelength groupings and the distribution thereof.

The invention takes short snapshots of an incoming signal (e.g. samples with a duration of 10 milliseconds) and keeps count of the number of occurrences of different wavelengths of the waveforms composing the incoming signal within a selected number of ranges. This method of sampling resembles the behavior of neurons of the human ear. For a given sample, voice will generate occurrences in a greater number of ranges than will tones because voice sounds exhibit more complex waveforms than tones which are composed of one or more of a few simple frequencies. For reliability, and to avoid an erroneous interpretation of noise, the invention can take two or three samples in succession such that the same successive interpretation will cause a response by the invention that distinguishes between silence, tones and voice on a telephone line.

For a better understanding of the invention, reference is made to the following description of an exemplary embodiment thereof, and to the accompanying drawings, wherein:

FIG. 1 is a block diagram of a voice detection system of the present invention;

FIG. 2 is a flow chart of a timer interrupt software program used by the voice detection system of FIG. 1; and

FIG. 3 is a flow chart of an analysis routine software program used by the voice detection system of FIG. 1.

As shown in FIG. 1, a plurality of subscriber telephone sets 11 are connected by cables 12 to a telephone switching network 13. The switching network 13, in turn, connects the sets 11, via respective telephone lines 14, to a plurality of hybrid circuits 15 of a telephone calling system (not shown). The hybrid circuits 15 interface the telephone switching network 13 with a voice detection system 16, which may form part of the telephone calling system, and separate incoming and outgoing signals that appear on the connected telephone lines 14. In this manner, only incoming signals will be applied to the voice detection system 16.

The voice detection system 16 comprises a plurality of zero crossing detectors 17, a plurality of latch elements 18 and a programmable logic element 19. Note that a power supply feeds the appropriate elements of the system 16. Each zero crossing detector 17 is tied to a respective telephone line 14 and detects when the incoming signal voltage on the telephone line 14 changes from a negative voltage to a positive voltage. The zero crossing detector 17 can detect when the incoming signal voltage changes from a positive voltage to a negative voltage since identical information would be obtained. For each so-called zero voltage crossing of the incoming signal voltage (i.e., for whichever polarity change is selected to be utilized), a voltage pulse is transmitted by the detector 17 to an associated latch element 18. At each of these zero voltage crossings, the length of the waveform of the incoming signal from the time of the previous crossing is measured by the logic element 19.

The latch elements 18 each comprise an input to receive the voltage pulses from the associated detector 17 and an output to produce a latched logic "1" in response thereto, thus storing the zero voltage crossing events. The programmable logic element 19, which may be a microprocessor or microcomputer, samples the output contents of the latch elements 18 and processes the frequency-related information for the respective incoming signals gathered by the system 16 according to a voice detection algorithm described below. The results of the information processing (i.e., determination of voice, call progress tones, silence, etc.) are passed on to the telephone calling system for further appropriate action.

In operation, the telephone calling system, of which the voice detection system 16 may be part, places telephone calls to a number of subscriber sets 11. The respective telephone lines 14 to the switching network 13 are then connected to the hybrid circuits 15 of the calling system. The hybrid circuits 15 pass only the incoming signals of the telephone lines 14 to the zero crossing detectors 17 of the voice detection system 16, each detector 17 being assigned to monitor one telephone line 14.

The incoming signal of each telephone line 14 may be composed of any number of waveforms of various wavelengths depending upon the condition of response on the telephone line 14 (i.e., silence, tones, voice, etc.). Each zero crossing detector 17 monitors the incoming signal of the assigned telephone line 14 and outputs a voltage pulse each time the incoming signal voltage changes from a negative to a positive polarity. Therefore, the detector 17 will not output a voltage pulse if silence is on the telephone line 14 and will output many voltage pulses if call progress tones or voice are on the line 14.

Each latch element 18 receives the output of a respective detector 17 and outputs a latched logic "1" upon receiving a voltage pulse. The logic element 19, which checks the output contents of the latch elements 18 simultaneously, is thus presented with a number of logic "1"'s equal to the number of telephone lines 14 that carry incoming signals. Upon recognizing that a zero voltage crossing has occurred for an incoming signal, the logic element 19 will issue a reset signal to the respective latch element 18 to prepare it for the next zero voltage crossing of the incoming signal.

FIG. 2 shows a flow chart of a timer interrupt software program which is utilized by the logic element 19 to sample the incoming signals on the telephone lines 14. The logic element 19 comprises a timer (not shown) which generates timer interrupts at a predetermined frequency (e.g., 3,000 hertz). The timer interrupts, which are set to be the highest priority interrupt for the logic element 19, prompt the timer interrupt routine (rectangle 20) to instruct the logic element 19 to read the output of the latch elements 18 (rectangle 21) and then to reset the latch elements 18 for the next zero voltage crossings of the respective incoming signals (rectangle 22).

For each latch element 18 utilized by the system 16 (diamond 23), the logic element 19 then tests if a zero voltage crossing event is stored therein (diamond 24). For those latch elements 18 that did not produce a latched logic "1" output, the logic element 19 goes on to test the next latch element 18 (diamond 23). For those latch elements 18 that produced a latched logic "1" output, the logic element 19 measures the elapsed time since the last crossing event by counting the number of interrupts between crossing events (rectangle 25). Counting the number of timer interrupts since the last zero voltage crossing permits the timing of incoming signal waveforms on a line 14 with a frequency greater than or equal to that of the timer (e.g., 3,000; 1,500; 1,000; 750; 600; 500; 429; 375; etc.). Finer gradations can be achieved by using a faster timer. Using the timing of the waveforms, the logic element 19 derives and establishes a registry of the wavelengths of the waveforms that compose the incoming signal on a line 14. As the crossing events are detected, the logic element 19 increments the count in the particular wavelength slots, or buckets, corresponding to the measured times between crossing events (rectangle 25). At the end of this operation, the next latch element 18 is tested in the same manner (diamond 23).

At the completion of testing of all the latch elements 18, the logic element 19 determines if the sampling time for the incoming signals (i.e., 10 milliseconds) has elapsed (diamond 26). If it hasn't, the logic element 19 returns to whatever task it was performing before the timer interrupt occurred (e.g. number-crunching, idling, etc). (oval 27). If it has elapsed, the logic element 19 stores the wavelength bucket information and initializes a new set of wavelength buckets for the next sampling period (rectangle 28). Next, the timer interrupt routine starts up an analysis routine to interpret the data collected during the sampling period (rectangle 29). The analysis routine, which is triggered by a second timer at, for example, 100 hertz, runs as a background task with respect to the timer interrupt routine, allowing the next samples to be stored without interference. When the analysis routine has completed its evaluation of the data, the logic element 19 again returns to whatever task it was performing before the timer interrupt occurred (oval 30).

FIG. 3 shows a flow chart for the analysis routine software program that is utilized by the logic element 19 to determine the condition of response on the telephone lines 14. The analysis routine (rectangle 35) is run for each of the telephone lines 14 being monitored, and, therefore, for each associated latch element 18 (diamond 36). As mentioned above, when the analysis routine has completed its evaluation of the data for each monitored telephone line 14, the wavelength buckets used by the logic element 19 are returned to free storage (rectangle 37) and the logic element 19 returns to its pre-timer interrupt task (oval 38).

The identical analysis is run by the analysis routine for the incoming signal of each telephone line 14 monitored. First, the logic element 19 determines if the total number of zero voltage crossings detected is below a predetermined threshhold limit (diamond 39). If the number is below the limit, then a temporary line state in the logic element 19 is interpreted as "silence" (rectangle 40). If a sufficient number of crossing exists, then the logic element 19 tests whether the incoming signal has a predetermined amount of wavelength buckets within a selected number of ranges (diamond 41). The ranges of wavelengths are derived from the frequency range of 428 to 3,000 hertz. If that amount of buckets are not "filled", i.e., do not have counts of zero voltage crossings, then the temporary line state is interpreted as "tone" (rectangle 42). If a sufficient amount of buckets are filled, then the temporary line state is interpreted as "voice" (rectangle 43). In a sense, the logic element 19 compares the stored wavelength bucket registry of the incoming signal with a model registry of an incoming signal composed of tones and proceeds accordingly.

Next, the logic element 19 compares the temporary line state with a current line state (rectangle 44). For the first sampling of a monitored telephone line 14, the current line state will be initially empty indicating no interpretation of the incoming signal. However, at the end of the first sampling, the logic element 19 will replace the contents of the current line state with the contents of the temporary line state. During the next sampling, a comparison can then be made between the two line states. If the temporary line state is different from the current line state and if the temporary line state has remained the same, for example, for three samples (i.e., 30 milliseconds), then the current line state will be changed to reflect the new state, i.e., the contents of the temporary line state replace the contents of the current line state (rectangle 44).

Upon the change to the new line state, the logic element 19 then examines the contents of the current line state. If the new line state is determined to be voice, this signals that the telephone line 14 has been answered and the logic element 19 can direct that other equipment of the calling system attached to the voice detection system 16 can respond accordingly. If the new line state has changed from silence to tone or from tone to silence, the logic element 19 can direct that the length of the previous interval be timed for decisions about ring or busy tones. When the analysis routine has completed the evaluation for a telephone line 14, the logic element 19 returns to the start of the routine (diamond 36).

As mentioned previously, finer gradations of sampling may be achieved by using a faster timer. "Polling" each zero-crossing detector 17 at a 3,000 hertz rate is based on the need to sample several detectors 17 with a single logic element 19 and the slow upper limit of frequencies of interest. In applications in which it is desired to sample for many discrete frequencies, or for higher frequencies, the necessity of a higher polling rate would change the economics of the application. For such an application, it is preferable to cause an interrupt with each zero voltage crossing and to measure the precise elapsed time from the last interrupt-zero voltage crossing. This technique can be used to cheaply fill as many wavelength buckets of so many discrete frequencies as were of interest, or alternatively, to allocate buckets of discrete frequencies as each new frequency range was experienced.

The embodiment described herein is merely illustrative of the principles of the present invention. Various modifications may be made thereto by persons skilled in the art without departing from the spirit or scope of the invention.

For example, although the above embodiment has an objective that requires only a crude distinction of frequency ranges, there is no limit to the number of frequency range buckets that can be filled for a finer analysis of sounds, not only for the analysis of voice, but also of speech or of music. The above embodiment also has no interest in capturing amplitude information of an incoming signal, but the invention can be modified to capture wave amplitudes as well as wavelengths.

Further, the invention can be used to sample and characterize sounds which do not originate within telephone or transmission lines. Ultrasonic generators for liquid atomization and flame hydrolysis torches for lightguide manufacture are examples of sound-generating devices which can be "fine-tuned" by using the invention to sample and characterize the sound signals generated by each.