WO1994023519A1

WO1994023519A1 - Method and apparatus for voice and modem signal discrimination

Info

Publication number: WO1994023519A1
Application number: PCT/US1994/002769
Authority: WO
Inventors: Huiyu Wang; Paul Odlyzko; Hungkun J. Chang
Original assignee: Motorola Inc.
Priority date: 1993-04-02
Filing date: 1994-03-16
Publication date: 1994-10-13

Abstract

A method and apparatus (232) for discriminating between voice and modem signals integrated into a single input signal operates upon the detection of a received signal's zero crossing rate and amplitude level in order to identify voice signals, data (modem) signals and silence.

Description

METHOD AND APPARATUS FOR VOICE AND MODEM SIGNAL

DISCRIMINATION

Field of the Invention

This invention pertains to a method' and apparatus for discriminating between voice and data (modem) signals as transmitted over a communications channel in a communication system. The present invention is particularly suited for use in a packet transmission system wherein the handling of burst- type voice and data packets typically require different signal processing techniques.

Background of the Invention

A current practice among telecommunications users is the transmission of data over the Public Switched Telephone Network (PSTN) utilizing data communications devices such as modems, facsimile devices, and other well known data carriers. Unfortunately, attractive methods for integrating the transmission of voice signals and data signals, hereinafter referred to as modem signals, via a common carrier are currently unavailable. This is due in part to the differing signaling techniques employed for voice and modem communications. Such differences include, but are not limited to the modulation of modem signals and the frequency bandwidth of voice signals.

The use of differing signaling techniques and error tolerances has lead to the adoption of differing and typically incompatible voice and modem signal error detection and correction techniques when voice and modem signal are transmitted as packets over digital transmission links. For example, voice packet substitution is a widely used error detection and correction technique for packetized voice communications. Voice packet substitution relies upon obtaining estimations of the pitch ^'(difference in the relative vibration frequency) of voiced speech. Such estimations are typically obtained via well known pulse code modulation (PCM) techniques. Whenever a voice packet is lost or incorrectly received, stored pitch estimations are substituted for the missing or corrupted speech, often with minimal interruption to the conversation.

For modem signal communications, packet retransmission, as opposed to packet substitution, has proven a more reliable and efficient error correction technique. Packet retransmission is typically characterized by a transmitted data packet having an error detection code (i.e., check sum,or cyclical redundancy check CRC) disposed within the transmitted signal. When received, the error detection code is examined to determine whether the received data packets has been corrupted. If so, the receiving device requests retransmission of the corrupted or missing packet. This retransmission protocol ensures that the data delivered to the user is error-free.

As will be appreciated by those skilled in the art, packet substitution, when utilized during the transmission of modem signals, fails to insure error- free data delivery. Packet substitution provides a best guess estimate of what the missing or corrupted information was. While this is an acceptable approach for voiced speech, it will nonetheless tend to corrupt packetized data by erroneously replacing bits within the data stream. In a similar fashion, the use of packet retransmission for voice communications introduces high delays and tends to lead to missing or choppy speech patterns which significantly degrades voice quality, thereby adversely impacting the overall performance of the system.

Current methods for discriminating voice and modem signals typically require some form of human intervention. Such techniques include, but are not limited to, manual switching, use of dual-tone multiple frequency (DTMF) signals generation, or the generation of distinct audible tones (rings) which indicate to the PSTN subscriber whether an incoming call comprises modem signals (e.g., fax) or voice (i.e., phone call). As a consequence, the individual desiring to transmit or receive both modem signals and voice signals must initiate two separate calls on separate telephone channels. This separation assures proper discrimination of voice and modem signals and simplifies the task of provisioning appropriate error detection and correction.

It would be extremely advantageous therefore to provide a method and apparatus for discriminating between voice signals and modem signals, thereby permitting the integrated transmission of both voice and modem signals over a common channel within a single call and with minimal human intervention.

Summary of Invention

Briefly described, the present invention is a method and apparatus for discriminating between voice and modem signals as disposed within an input signal. The present invention sets forth method steps and apparatus structure for determining, during a predetermined interval, a number of polarity changes (zero crossings) within an input signal comprising voice and modem signals, measuring the s'ignal's amplitude during the predetermined interval, and detecting the presence of modem signals and voice signals disposed within the input signal as a function of the number of polarity changes and the signal amplitude.

In accordance with one embodiment, the discriminator circuit of the present invention is employed within a telephone I/O device in order to enable the appropriate error detection and correction circuitry during the reception of voice signals and modem signals, respectively.

In accordance with another embodiment, the discriminator circuit of the present invention is employed within a switching circuit in order to detect the presence of voice signals, modem signals and silence within a received input signal. Detected voice signals are routed to a voice instrument, while detected modem signals are routed to a modem device. The modem device is operated to transmit modem signals during the detection of silence within the received input signal.

Brief Description of the Drawings

FIG. 1 illustrates a wireless Local Area Network (LAN) suited for incorporating the present invention; FIG. 2 illustrates a block diagram of a packet transmission device for incorporating the present invention;

FIG. 3 illustrates a statistical distribution of zero crossing rates for voiced and unvoiced speech signals;

FIG. 4 is a table of information providing zero crossing rates for modem signal modulation schemes;

FIG. 5 is a detailed circuit diagram of a discriminator circuit in accordance with the present invention;

FIG. 6 is a block diagram embodiment of a switching circuit which employs the discriminator circuit of the present invention; FIG. 7 is another block diagram embodiment of a switching circuit that employs the discriminator circuit of the present invention;

FIG. 8 is a flow chart diagram of the steps performed by the discriminator circuit of the present invention for detecting zero crossing rates;

FIG. 9 is a flow chart diagram of the steps performed by the discriminator circuit of the present invention for detecting signal amplitudes; and

FIG. 10 is a flow chart diagram of the steps performed by the discriminator circuit of the present invention for differentiating between voice and modem signals.

Detailed Description

With reference to the present invention, the zero crossing rate (ZCR) of a received signal wave form is determined in order to distinguish voice signals from modem signals. Unfortunately, ZCR alone is inadequate to achieve this objective. In accordance, the wave form signal's amplitude in conjunction with its ZCR is determined in order to accurately distinguish voice signals from modem signals as transmitted over a communications line. FIG. 1 illustrates a wireless local 'area network (LAN) suited for incorporating the present invention. In the illustrative network 100 a control module (CM) 110 utilizes radio frequency (RF) communications to communicate with user modules (UM) 120 that are each coupled to user devices 140 consisting of a terminal, personal computer, telephone, or other information input/output device. CM 110 is coupled by communications channel 160 and central switching office (CSO) 180 to an external voice and data communications system, such as the PSTN, a radiotelephone network (RTN) or any of the other well known voice/data networks. As will be appreciated, channel 160 may be implemented via_,any of the well known voice and data communication channel media, such as, but not limited to wire, optical links and radio links.

CM 110 controls communications within the illustrative network and passes information from the PSTN via CSO 180 and channel 160 to user devices 140 via an associated UM 120. CM 110 also controls local communications by receiving information from one UM 120 and relaying the information to a different UM 120. The information conveyed between CM 110 and UMs 120 is transmitted in the form of packets. While the system of FIG. 1 utilizes RF technology for CM to UM communications, it will be appreciated by those skilled in the art that any available wireless or wire line communications media may be substituted in place of the RF communications equipment described herein without departing from the spirit of the present invention.

FIG. 2 illustrates a block diagram of a packet transmission device which includes the present invention. The illustrative embodiment is for a UM 120 of the system 100 as shown in FIG. 1. A communications controller 200 includes a microprocessor (MPU) 218, with associated read only memory (ROM) 210, random access memory (RAM) 212, and a network interface 214. The MPU 218 operates under the direction and control of an operating instruction set stored in ROM 210 in order to control communications between UM peripheral devices 222 and 228 and CM 110. Network interface 214 consists of appropriate registers and line drivers for communication with those peripheral devices connected to the network interface (NI) bus 220. The physical structure of the NI bus 220 is known in the art,and therefore requires no additional discussion at this time.

In the structure as shown in FIG. 2 a plurality of peripheral devices including two way RF radio 222 and a telephone I/O device 228 are connected to the communications controller 200. The illustrative peripherals are merely representative that virtually any type of packetized information can be coupled by means of an appropriate input/output device to UM 120. The peripherals each contain a NI bus interface 226 and 230 respectively. These interfaces provide the necessary registers and line drivers for communicating with NI bus 220 and will also include an MPU, RAM, and ROM if these resources are not available in the peripheral device. The radio 222 includes one or more antennas 224 for RF communications with CM 110 as shown in FIG. 1. The telephone I/O device 228 is connected by terminal 236 to a telephone instrument and a modem device (not shown) . In accordance with the present invention, the telephone I/O device 228 employs a discriminator circuit 232. The function of the discriminator circuit 232 is to distinguish voice signals from modem signals as received from the NI bus 220. Thereafter, the discriminator circuit 232 will control the operation of voice substitution circuit 234. When discriminator circuit 232 identifies voice signals, it enables voice substitution circuit 234 which provides error detection and correction during the reception of voice signals. When discriminator circuit 232 identifies modem signals, voice substitution circuit 234 is disabled such that packet retransmission techniques, as known in the art, may be employed to provide error detection and correction during the transmission and reception of modem signals.

While the preferred embodiment shows a NI bus 220 connecting the various peripherals to the communication controller 200, it will be appreciated by those skilled in the art that the NI bus can be substituted by a time-division-multiplexed (TDM) bus, bi-directional buses or packet switches which are all known in the art. Of note, CM 110 has a configuration substantially identical to that shown in FIG. 2.

During UM 120 operation, packetized information (voice and/or modem signals) is received by radio 222. Radio 222 contains a conventional set of equipment for providing radio coverage including conventional components to perform the functions of a transmitter and a receiver (i.e., a transceiver). The received information is processed by the conventional radio circuitry, and routed to the communications controller 200 for storage in RAM 212. Thereafter, the communications controller 200 passes the packetized information to the appropriate peripheral via NI bus 220. Communications controller 200 and NI bus 220 operation is known in the art, and requires no additional discussion at this time. The interested reader may nonetheless refer to US Patent Application No. 07/ 719,212 filed 06/21/90 and assigned to the assignee of the present application. It is an aspect of the present invention for the discriminator circuit 232 of FIG. 2 to distinguish voice signals from modem signals. Due to the relatively slow rate of change of resonant cavities exhibited over 40 milliseconds of speech, speech signals exhibit considerable informational redundancy over intervals of tens of milliseconds or less. When limited to the conventionally accepted voice communication frequency range of 300 Hertz (Hz) to 3.5 kilohertz (K Hz), speech signals are typically categorized as being of two types: voiced speech and unvoiced speech.

Unvoiced speech is produced by turbulence at points of constriction in the vocal tract and is typically characterized by low amplitude and frequent sign changes. Voiced speech is produced by vibration of vocal cords and is typically characterized by high amplitude and infrequent sign changes. The summation of the number of sign (polarity) changes that occur to a wave form over a predetermined sampling interval is often referred to as the zero crossing rate. As is known in the art, ascertaining a signal's zero crossing rate (ZCR) is a technique typically employed to estimate a wave form's frequency. A zero crossing is said to occur when successive wave form samples have different algebraic signs (e.g., the wave form crosses the zero axis) .

FIG. 3 illustrates a statistical distribution of the ZCR for voiced and unvoiced speech. As shown, and known in the art, voiced and unvoiced speech are ^■ distinguishable by evaluating the number of polarity changes that occur within a wave form over a predetermined sampling interval. By comparing these values to predetermined thresholds, voiced and unvoiced speech signals can be distinguished. For unvoiced speech, the average ZCR is approximately 29 zero crossings per 4 millisecond interval. For voiced speech, the average ZCR is approximately 7-8 zero crossings per 4 millisecond interval. The adopted 4 millisecond sampling interval is merely illustrative, thus sampling intervals of varying duration may be utilized without departing from the spirit of the present invention.

While the use of ZCR to distinguish voice from unvoiced speech is known, the present invention recognizes that ZCR distributions for voice and modem signals tend to overlap. As a result, it is not always possible to distinguish voice signals from modem signals based upon zero crossing rates alone. FIG. 4 is a table providing carrier frequency, data rate, and zero crossing rate information for various modem signal modulation schemes. The modulation schemes considered in FIG. 4 comprise Differential Phase Shift Keying (DPSK) , Frequency Shift Keying (FSK) , Trellis-Coded Modulation (TCM) , and Quadrature Amplitude Modulation (QAM) . While other modulation techniques, such as, but not limited to Amplitude Phase Keying (APK) , Amplitude Shift Keying (ASK) , Binary Phase Shift Keying (BPSK) , Multiple Phase Shift Keying (MPSK) and Quadrature Shift Keying (QPSK) are available, the listing provided in Fig. 4 is considered representative.

Modem signals are typically characterized as narrow band signals, thus they tend to exhibit a relatively high ZCR.

Upon review of FIG. 4 the average zero crossing rate (ZCR) for the representative modem signals is greater than 8 zero crossings per 4 millisecond sampling interval. From this information it should be appreciated that examination of ZCR alone is insufficient to distinguish voice signals from modem signals. The average ZCR for voiced speech is approximately 7-8 zero crossings per 4 milliseconds and 29 zero crossings per 4 millisecond interval for unvoiced speech.

In order to overcome this problem, the present invention, monitors signal amplitude along with ZCR to enable the precise differentiation between voice signals (voiced and unvoiced speech) and modem signals.

Based upon observation it has been determined that modem signals may be fairly characterized as signals having a relatively high amplitude and ZCR (8-15 zero crossings every 4 milliseconds) . Voice signals, consisting of voiced and unvoiced speech, may be characterized as follows:

- Voiced speech comprises signals having a relatively high amplitude and low ZCR (4-8 zero crossings every 4 milliseconds) ; - Unvoiced speech comprises signals having a relatively low amplitude which is equivalent to 10-25 percent of the average voiced speech amplitude and high ZCR (more than 15 zero crossings every 4 milliseconds) .

Based upon these observed differentiations, it is now possible to distinguish voice signals from modem signals by determining a received wave form's ZCR and amplitude and comparing the exhibited characteristics to predetermined threshold values. FIG. 5 is a detailed circuit diagram of the discriminator circuit 232 of FIG. 2. In the illustrative embodiment, an input terminal^'500 receives PCM information representative of sequential samples in a voice/data packet. The PCM samples are input into a packet zero crossing counter (PZCR) 502 which increments by one when the most significant bit (MSB) of the current PCM sample is different from that received in a previous PCM sample. When a_, 4 millisecond (msec) time period has elapsed a packet end signal 501 is generated by the communications controller 200 of FIG. 2 which enables comparator 504. The comparator 504 compares the result of counter 502 to a predetermined value loaded in the register 506. This value is equal to ((cut-off frequency of voiced speech * 2)/1000) * (N/8) ) , where N is the total number of samples obtained during a sampling interval. In accordance with the present invention N is 32 and the cut-off frequency of voiced speech is approximately 900 Hz. In accordance the value stored in register 506 is 8. If the count of counter 502 is less than the threshold packet zero crossing (TPZC) value maintained in register 506, comparator 504 outputs a logic 1 to logic circuit 518. If the count of counter 502 is greater than or equal to the value in register 506, comparator 504 outputs a logic 0 to logic circuit 518. Once the comparison has been completed, packet end signal 501 resets PZCR counter 502.

In conjunction with the above description, PCM samples are simultaneously input to comparator 508 and compared to a predetermined valued stored in threshold amplitude (TAMP) register 510. For each sample that equals or exceeds the predetermined threshold TAMP, comparator 508 outputs a logic 0 to the counter 512. For each sample that is less than TAMP, comparator 508 outputs a logic 1 to total counter 512. Counter 512 increments by one each time a logic 1 is received from comparator 508. When a 4 msec time period has elapsed, a packet end signal 501 is generated by the communications controller 200 of FIG. 2, which enables comparator 514. Comparator 514 then compares the result of counter 512 to a predetermined value stored in the register 516. This value is equal to 90 percent of the total number of PCM samples per packet (N * 0.90). If counter 512 count is less than the predetermined value stored in register 516, comparator 514 outputs a logic 0 to logic circuit 518. If, on the other hand, counter 512 count is greater than or equal to the value stored in register 516, comparator 514 outputs a logic 1 to logic circuit 518. Once the comparison has been completed the packet end signal 501 resets counter 512. The results of comparator 504 and comparator 514 are input to logic circuit 518 which determines if the signal comprises voice signals. (i.e., voiced or unvoiced speech), modem signals or silence.

Silence is detected by logic circuit 518 when comparators 504 and 514 both output logic 1 values, respectively. Voiced speech is identified by logic circuit 518 when comparators 504 and 514 output a logic 1 and a logic 0, respectively. Unvoiced speech is identified by logic circuit 518 when comparators 504 and 514 output a logic 0 and a logic 1, respectively. Modem signals are thus identified when comparators 504 and 514 both output logic 0 values, respectively.

In the illustrative embodiments of the present invention, the automatic detection of voice signals, modem signals and silence offers the advantage of eliminating unnecessary and disruptive delays in telecommunication systems that must support transmission of the above information, thereby improving overall system throughput. The present invention also overcomes the need to restrict modem and voice signal transmissions to separate telephone calls. In addition, human intervention during voice and modem signal discrimination is reduced to a minimum. FIG. 6 is a block diagram of an embodiment of a switching circuit 600 that employs the discriminator circuit of the present invention. Switching circuit 600 is designed to receive an analog input signal 602 comprising both voice and modem signals and to thereafter distinguish one from another. As shown, the switching circuit 600 comprises an analog-to- digital converter (A/D) 610, discriminator circuit 620 and a switch 630. Switch 630 couples the device 600 to a modem device 640 via line 645. Modem 640 operates as is known in the art with a data terminal 650 such as a photocopy machine, facsimile machine, printer, personal computer or any other data transmission/reception device to modulate and demodulate received signals. In addition, a telephone instrument 660 is connected to'the switch 630 via line 645.

During operation, the analog input signal 602 is converted by A/D converter 610 into numerous digital representations which are routed to discriminator circuit 620 which detects the presence of voice signals, modem signals or silence.

In accordance with the preferred embodiment, A/D 610 operates at an 8 K Hz sampling rate (8000 samples per second). Thus, every 4 milliseconds, 32 samples (a packet) of the analog signal are taken. Each sample is converted into an 8 bit digital representation. Discriminator circuit 620 operates upon the received digital information to determine a relative ZCR and signal amplitude. The exhibited signal characteristics are then compared to predetermined thresholds to identify the presence or absence of voice and/or modem signals. Based upon the comparison, discriminator circuit 620 controls the operation of switch 630 which routes the analog input signal 602 to the modem 640 or the telephone instrument 660 depending upon the outcome of the comparison. The switch 630 is an analog switch like those known in the art. In the past, such switches have been available under part number MC54/74HC4053 by contacting contacting Motorola Inc., through its Semiconductor Products Sector at 2200 West Broadway, Mesa, Arizona 85201. In accordance with a preferred embodiment, discriminator circuit 620 is implemented via the structure of FIG. 5. In accordance with yet another embodiment, discriminator circuit 620 may be implemented as a programmable digital signal processor (DSP) , programmed in accordance' with instruction provided herein below. Such a DSP has in the past been available under part number DSP56156 or DSP56156ROM by contacting Motorola Inc., through its - Semiconductor Products Sector at 2200 West Broadway, Mesa, Arizona 85201.

While A/D 610 and discriminator circuit 620 are depicted as separate devices, it will be appreciated by those skilled in the art that the functionality of these devices may be integrated into a single device, such as, for example the DSP56156 or DSP56156ROM as previously mentioned. It will also be appreciated by those skilled in the art that switching circuit 600 may be implemented as an Application Specific Integrated Circuit (ASIC) designed to operate in accordance instruction provided herein below.

FIG. 7 is another block diagram embodiment of a switching circuit 700 that employs the discriminator circuit of the present invention. Switching circuit 700 operates substantially in accordance with the circuit described in association with FIG. 6, except the switching circuit 700 receives a digital input signal 702, which is operated upon by discriminator circuit 710 in accordance instruction provided herein below to determine the presence of modem signal, voice signals or silence. Thereafter, digital-to- analog converter (D/A) 720 is employed to convert the digital input signal into an analog output signal. The analog output signal is then routed to modem 740 or telephone device 760 via the discriminator controlled switch 730. As will be appreciated, switching circuit 700 may be implemented via the discriminator circuit 500 of FIG. 5, a programmable DSP, or an Application Specific Integrated Circuit (ASIC) designed to operate in accordance 'instruction provided herein below.

FIG. 8 is a flow chart diagram of the steps performed by the discriminator circuit of FIGS. 6 and 7 for evaluating the ZCR of a received signal assuming said discriminator circuit is implemented as a DSP. Beginning with START block 800, flow proceeds to block 802 where a zero crossing rate variable or counter ZCR and a sample number variable or counter i are set to zero. At block 804 a check is performed to determine whether the polarity of each sample within a packet comprising 32 samples has changed. If the polarity of the current sample, as determined by examining its most significant bit (MSB), is different from that of the previously received sample, flow proceeds to block 806 where the zero crossing rate variable or counter ZCR is incremented. At block 808 the sample variable or counter i is incremented when no sign change occurred at block 704 or when ZCR is incremented at block 806.

At decision block 810 a check is performed to determine whether all 32 (N) samples comprising the current packet have been evaluated for ZCR. If not, flow branches back to block 804 where a next sample is evaluated. When all 32 samples have been considered, flow proceeds to decision block 812 where the ZCR variable or count is compared to a threshold value to determine whether the evaluated packet has a HI or LOW ZCR. From the observations set forth above, when the ZCR variable or count is less than or equal to a threshold zero crossing of 8, the packet meets those conditions necessary to receive a LOW zero crossing rate status. Such LOW status is indicated at block 814. Conversely , when the ZCR variable or count is greater than the threshold zero crossing (TZC) of 8, the packet meets the condition necessary to receive a HI zero crossing rate status. Such HI status is indicated at block 816.

FIG. 9 is a flow chart diagram of the steps performed by the discriminator circuit of FIGS. 6 and 7 for evaluating the signal amplitude of a received signal assuming said discriminator circuit is implemented as a DSP. Commencing with START block 900 flow proceeds to block 902 where an amplitude variable or counter AMP and a sample number variable or counter i are set to zero. In accordance with the preferred embodiment the AMP variable tracks the total number of voice samples in a packet having an amplitude less than a predetermined threshold. It will nonetheless be appreciated by those skilled in the art that the AMP variable could track the total number of samples in a packet having an amplitude greater than the amplitude threshold without departing from the spirit of the present invention.

At block 904, a check is performed to determine whether the absolute value of a sample's amplitude is less than a predetermined amplitude threshold. If so, flow proceeds to block 906 where the AMP variable in incremented. In accordance with the preferred embodiment, the MSB of each sample represents signal polarity while the remaining bits represent amplitude. In accordance, amplitude is readily obtainable from the digital bit representations.

At block 908 the sample variable or counter i is incremented when the sample amplitude exceeded the threshold at block 904 or when AMP is incremented at block 906. At decision block 910 a check^' is performed to determine whether all 32 (N) samples comprising the current packet have been evaluated for amplitude. If not, flow branches back to block 904 where a next sample is evaluated. When all 32 samples have been considered, flow proceeds to decision block 912 where a check is performed to determine whether the AMP variable count exceeds a threshold determined by N * 0.9. When 90 percent of the evaluated samples comprising the current packet have an amplitude less than the predetermined threshold, the packet is deemed to have a low amplitude. Under this condition, a LOW amplitude status is indicated at block 914. When fewer than 90 percent of the evaluated packet samples have an amplitude less than the threshold, the packet is deemed to have a high amplitude. Under this condition, a HI amplitude status is indicated at block 916.

Having identified salient characteristics of voice and modem signals, FIG. 10 illustrates a flow chart diagram of the steps performed by the discriminator circuit of the present invention for discriminating between voice signals and modem signals. Commencing at START block 1000, upon completion of the flow process described in association with FIGS. 8 and 9, flow proceeds to decision to block 1010 where a check is performed to determine whether the packet ZCR status of a received packet, as determined by step 812 of FIG. 8, is HI. If so, flow proceeds to decision block 1020 where a check is performed to determine whether the packet amplitude status of the received packet, as determined by step 912 of FIG. 9, is HI or LOW. If the amplitude status is HI, the discriminator circuit will identify the current packet as comprising modem signals and take appropriate action consistent with the processing of modem signals at block 1030. If, on the other hand, the packet amplitude is determined to be LOW, at block 1020, the discriminator will identify the current packet as comprising unvoiced speech type voice signals and thereafter take appropriate action consistent with the processing of unvoiced speech signals at block 1040. Assuming the packet ZCR is determined at block 1010 to be LOW, flow will proceed to decision block 1050 where a check, identical to that performed at block 1020, is performed to determine whether the packet amplitude is HI or LOW. If the amplitude is HI, the discriminator circuit will identify the current packet as comprising voiced speech type voice signals and thereafter take appropriate action consistent with the processing of voiced speech signals at block 1060. If, on the other hand, the packet amplitude is determined to be LOW, at block 1050, the discriminator will identify the current packet as comprising silence and thereafter take appropriate action consistent with the processing of silence at block 1070. It will be appreciated by those skilled in the art that the type of post discrimination activity considered by the current invention includes, but is by no means limited to: routing detected modem signals to modem devices, routing detected voice signals to a telephone instruments and other voice processing devices. Moreover, the present invention anticipates utilizing the presence of silence as detected in the input signal, to prompt a modem device to begin the transmission of modem signals. Such a prompt may be generated by the discriminator circuit. In accordance with the present invention, it is anticipated that modem device data transmissions will continue until voice and/or modem signals are detected within the received input signal. Such detection will operate to inhibit modem device data transmissions. In this way, bi¬ directional voice and modem signal transmissions may be multiplexed on a common communications line, simultaneously.

In accordance with the present invention, the automatic detection of voice signals, modem signals and silence offers the advantage of eliminating unnecessary and disruptive delays in telecommunication systems that support transmission of voice and data. The present invention also overcomes the need to restrict modem and voice signal transmissions to separate calls. In addition, human intervention during the discrimination process is reduced to a minimum.

Claims

1. A method for discriminating voice signals from mode signals disposed within an input sign,al comprising the steps of: determining, during a predetermined interval, a number of polarity changes within the input signal; measuring the input signal's amplitude during the predetermined interval; and detecting a presence of modem signals and voice signals disposed within the input signal as a function of the number of polarity changes and the amplitude of the input signal.

2. The method of claim 1 further comprising the steps of detecting the presence of modem signals within the input signal when both the signal amplitude and the number of polarity changes exceed respective thresholds.

3. The method of claim 1 further comprising the steps of: identifying voice signals within the received input signal when the amplitude exceeds an amplitude threshold and the number of polarity changes does not exceed a count threshold; and identifying voice signals within the received input signal when the amplitude does not exceed the amplitude threshold and the number of polarity changes exceeds the count threshold.

4. The method of claim 1 further comprising the step of detecting the presence of silence within the input signal when the signal amplitude does not exceed the amplitude threshold and the number of polarity changes does not exceed the count threshold.

5. The method of claim 1 wherein the input signal comprises signals selected from the group consisting of analog signals, pulse code modulated representations of analog signals, and digital signals .

6. A method for discriminating voice signals from modem signals wherein said voice and modem signals are multiplexed within an input signal, said method comprising the steps of: receiving an input signal comprising voice and modem signals; during a predetermined interval, sampling the received input signal to acquire N samples; determining a number of zero crossings within the sampled signal as a function of the acquired samples; determining the sampled signal's amplitude as a function of the acquired samples; and detecting a presence of modem signals and voice signals within the input signal as a function of the zero crossing count and the amplitude of the sampled signal.

7. The method of claim 6 further comprising the steps of: identifying modem signals within the received input signal when the sampled signal's amplitude exceeds an amplitude threshold and the zero, crossing count exceeds a zero crossing count threshold; identifying voice signals within the received input signal when the sampled signal's amplitude exceeds the amplitude threshold and the zero crossing count does not exceed the zero crossing count threshold; identifying voice signals within the received input signal when the sampled signal's amplitude does not exceed the amplitude threshold and the zero crossing count exceeds the zero crossing count threshold; and identifying silence within the received input signal when the sampled signal's amplitude does not exceed the amplitude threshold and the zero crossing count does not exceed the zero crossing count threshold.

8. The method of claim 6 further comprising the steps of: routing detected voice signals to a voice instrument; routing detected modem signals to a modem device; and transmitting modem signals from the modem device when silence is detected within the input signal. 9. The method of claim 7 wherein the amplitude threshold is defined as N * 0.

9, where N is the number of samples acquired during the predetermined interval.

10. The method of claim 7 wherein the zero crossing threshold is defined as ( (cut-off frequency * 2)/1000) * (N/8) ) , where N is the number of samples acquired during the predetermined interval .

11. A switching circuit for discriminating voice signals from modem signals wherein said voice and modem signals are multiplexed into a single input signal, the switching circuit comprising: means for receiving the input signal; circuit means, coupled to the receiving means, for processing the received input signal to determine a number of zero crossings and an input signal amplitude exhibited during a predetermined interval; and a detector coupled to the circuit means, for detecting a presence of voice signals, modem signals and silence within the input signal as a function of the zero crossing count and the input signal amplitude.

12. The circuit of claim 11 further comprising a switch coupled to the detector for routing modem signals and voice signal to respective destinations.

13. The circuit of claim 11 further comprising means for sampling the analog input signal at an 8K sampling rate over 4 millisecond intervals to acquire 32 signal samples.

14. The circuit of claim 13 wherein each signal sample is converted into a digital representation.

15. The circuit of claim 14 wherein the most significant bit of the digital representation comprises signal polarity information and the remaining bits of the digital representation comprises signal amplitude information.