WO1996014694A1

WO1996014694A1 - Method and apparatus for establishing full-duplex sound communication for a speakerphone system

Info

Publication number: WO1996014694A1
Application number: PCT/US1995/014530
Authority: WO
Inventors: Chang-Tsuo Liu; David K. Long
Original assignee: Sierra Semiconductor Corporation
Priority date: 1994-11-04
Filing date: 1995-11-03
Publication date: 1996-05-17

Abstract

The present invention is directed to an apparatus (such as a speakerphone (100)) and method for establishing a full-duplex signal path immediately upon connection of the apparatus to an external communication path (102) (such as an external telephone line). Exemplary embodiments of the present invention are directed to training the electrical echo canceler (138) with respect to the impedance of the external communication path (102) (such as an external telephone line to which the speakerphone (100) is connected) during an installation phase and storing the trained parameters for subsequent use prior to interfacing the system to an external telephone line and completing a full-duplex path connection.

Description

METHOD AND APPARATUS FOR ESTABLISHING FULL-DUPLEX SOUND COMMUNICATION FOR A SPEAKERPHONE SYSTEM

BACKGROUND OF THE INVENTION

Field of the Invention;

The present invention is generally directed to methods and apparatus for sound (e.g. voice) communication, and more particularly, to methods and apparatus for establishing and maintaining full-duplex signal paths for sound communication over, for example, a two-wire switched telephone network of a speakerphone system.

State of the Art:

Speakerphone systems which can provide full-duplex signal paths are known. For example, speakerphone systems are known wherein a microphone and a loud speaker are formed in an integral unit for connection with a two- wire switched telephone network. In the past, such speakerphones were typically configured for half-duplex operation; that is, the speakerphone was voice-switched to permit, at any given time, signal communication in only one direction over an external telephone line connected to the speakerphone. More recently, full-duplex speakerphones have been developed which include an electrical echo canceler (EEC) and an acoustic echo canceler (AEC). For example, a document entitled "HVP-S Speaker-Trained Voice Recognizer with Full- Duplex Speakerphone for Telecommunications Terminals" available from AT&T Microelectronics, December, 1993, provides a functional description of a speakerphone having a full-duplex mode of operation, and which includes both acoustic echo cancellation and voice-adaptive electrical echo cancellation. A catalog entitled "Hello Direct", Fall, 1994, describes a speakerphone which is also capable of full-duplex operation, available from Polycom.

Acoustic echoing occurs when sound outputs from the speaker of a speakerphone are received by a microphone of the speakerphone. In a full-duplex mode of operation, both the microphone and speaker are maintained active, such that this feedback path can return the sound output from the speaker to the telephone line and back to the far end of the telephone line where the sound originated.

Electrical echos occur at the telephone line interface where the two wire telephone line circuit is converted to the four wire speakerphone circuit. Via the line interface, sound received by the microphone is partially returned to a receiver circuit of the speakerphone and reaches the speaker. This is also known as sidetone. Further, electrical echoes occur within the telephone line circuit where impedance mismatches occur.

The combined effects of both electrical and acoustic echos results in sounds which can circulate around the loop from microphone to line interface to speaker and back to the microphone. If there is sufficient amplification in this loop to overcome the echo loss, the system will become regenerative and will oscillate causing a loud whistle or singing noise.

Accordingly, full-duplex speakerphone systems typically include acoustic echo cancellation and electrical echo cancellation to prevent regeneration and distortion which can occur due to these undesired acoustic and electrical feedback paths. The circuitry for providing acoustic echo cancellation is typically adapted, or trained, to account for: (1) the specific speakerphone in which it is used (that is, to account for the spacing of the microphone relative to the speaker in the speakerphone); and (2) the surrounding environment in which the speakerphone has been placed (for example, to account for the configuration of a room in which the speakerphone is placed such that reflected sound waves from the speaker output will not result in undesired acoustic feedback). Similarly, the electrical echo canceler of the speakerphone can be adjusted to reduce electrical echoing at the telephone line interface. The acoustic echo cancellation is dependent on relative placement between the microphone and speaker, and on the surrounding environment in which the speakerphone is placed. The canceler is typically pre-trained during an initialization period. This initialization occurs prior to the establishment of a telephone line connection between the speakerphone and an external telephone (that is, after the speakerphone has been powered up, but before a number has been dialed into the speakerphone, and a connection over a telephone line has been established). More particularly, during an initial mode of operation, a sound, such as a tone, is emitted from a speakerphone. The emitted tone is detected by the microphone, and coefficients of the acoustic echo canceler are trained to account for the acoustic feedback path established by the relative placement of the microphone and speaker, and by the surrounding environment.

Such techniques for cancelling acoustic echoes may be insufficient to provide sufficient system margin for full-duplex operation. In addition, known techniques for initializing acoustic echo cancellation are reliant upon the relative microphone/speaker positioning and the surrounding environment. Any subsequent relative movement between the microphone and speaker, or change in the surrounding environment (such as moving the microphone or speaker or changing characteristics of the environment by, for example, opening a door or window) can degrade system performance and can result in an annoying singing from the speaker or over the telephone line.

While this initialization training process reduces the acoustic echo it is typically too short a time to attain maximum effect. Attaining less than the maximum effect during the initialization process has been deemed acceptable since full training typically takes tens of seconds, and consumers will not accept a product that emits a sound for over 30 seconds every time it is turned on. Therefore, the initial training is a compromise, and the acoustic echo cancellation circuit is usually further trained or adjusted during the course of a telephone conversation until the optimum cancellation is achieved.

The electrical echo cancellation circuit for the speakerphone is trained after the communication path is established. Prior to making a telephone line connection, the electrical echo cancellation circuit (EECC) is typically preset to a nominal condition. After the call is established the electrical echo cancellation circuit can be trained to improve the cancellation effectiveness.

Because the acoustic and electrical echo cancelers are not fully trained at the beginning of a telephone call, they cannot be relied upon to provide sufficient echo reduction to avoid the singing effect described above. Accordingly, during an initial period at the beginning of a telephone call and at other times when the echo cancellation is ineffective, it is common practice to reduce the signal amplification in one direction of the signal path to avoid regeneration. The reduced signal strength in this path effectively cuts off the communication in that direction. This is known as a half-duplex mode. Typical speakerphones sense the direction of the strongest signal and cut back the amplification in the other path. After the echo cancelers are trained, full-duplex operation is enabled. Until this occurs, the annoying effects of half-duplex signaling must be endured by both parties to the conversation.

Accordingly, it would be desirable to provide a speakerphone capable of immediate full-duplex operation which eliminates any such need for the relatively slow and annoying, half-duplex initialization procedure of an acoustic echo canceler.

SUMMARY OF THE INVENTION

The present invention is directed to an apparatus (such as a speakerphone) and method for establishing a full-duplex signal path immediately upon connection of the apparatus to an external communication path (such as an external telephone line). Exemplary embodiments of the present invention are directed to training an electrical echo canceler with respect to the impedance of the external communication path (such as a telephone line to which the speakerphone is connected) prior to using the system to place a call on the telephone line.

The present invention takes advantage of three important facts concerning the electrical echo path. First, the dominant component of the echo is caused by a mismatch of the terminating impedance in the telephone's two-to-four wire circuit converter (commonly referred to as a "hybrid") and the impedance of the telephone line local loop (that is, the physical line connecting a subscriber to a central office or private branch exchange (PBX)). Secondly, the physical connection established between the hybrid and the telephone line is the same for every call made on that line. Third, the impedance of the telephone line does not change significantly between the time the telephone goes off hook and breaks dial tone, and the time the connection is established to the called party or calling party.

Exemplary embodiments of the invention have an initialization phase which is performed at the time of installation of the speakerphone and upon subsequent relocation of the speakerphone. In the initialization process, the microphone and speaker are turned off while the speakerphone is taken off hook and dual tone multiple frequency (DTMF) digits are transmitted. The number of digits transmitted (for example, two digits) is sufficient to break a first dial tone and, if using a PBX, a second dial tone and to connect the speakerphone to the central office, but not sufficient to establish a connection to a second party. At this time the line impedance presented to the speakerphone is virtually the same as would be present during a communication session. A locally stored training signal is then sent to the speakerphone two-to-four wire circuit and the EECC is trained. The coefficients thus derived are stored and used to preset the EECC at the outset of subsequent calls. In accordance with exemplary embodiments, significant advantages are realized. By reducing or eliminating electrical echos due to impedance mismatches between the speakerphone and the telephone line to which it is connected, loop stability can be improved to a point where full-duplex operation can be initiated immediately at the time communication is established over the telephone line. Because adequate echo cancellation can be achieved due to pre-training of the electrical echo canceler without dependence on pre-training of the acoustic echo canceler, relative movement between the microphone and speaker of the speakerphone, or changes in the environment, will not affect the ability to provide full-duplex operation. In addition, pre-training of the electrical echo canceler can be performed relatively quickly. That is, because the signal path delays associated with the electrical echo canceler are substantially shorter than those of an acoustic echo canceler (for example, sound waves travel much slower than electrical signals), pre- training of the electrical echo canceler can be achieved rapidly. Exemplary embodiments which encompass the foregoing features relate to a method and apparatus for establishing a full-duplex signal path comprising means for transducing sound waves into electrical signals; means for transducing electrical signals into sound waves; means for interfacing said sound wave transducing means and said electrical signal transducing means with an output of the apparatus, said interfacing means further including means for reducing electrical echoing within said interfacing means, and means for adaptively controlling operating characteristics of said electrical echoing reducing means prior to activating said output.

BRIEF DESCRIPTION OF THE DRAWINGS The present invention can be further understood with reference to the following description and the appended drawings, wherein like elements are provided with the same reference numerals. In the drawings:

Figure 1 illustrates an exemplary embodiment of an apparatus in accordance with the present invention; and Figure 2 is an alternate exemplary embodiment of an apparatus in accordance with the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Figure 1 shows an exemplary embodiment of an apparatus configured as a speakerphone 100 in accordance with an exemplary embodiment of the present invention. The speakerphone 100 can establish a full-duplex signal path over a communication path 102, represented as a two- wire telephone line output which can connect with a two-wire switched telephone network.

The Figure 1 speakerphone generally includes a means 104 for transducing sound waves into electrical signals using, for example, a transducer such as microphone 106. The microphone 106 can be any standard microphone readily available for converting input sound waves, such as voice communication, into electrical signals. The speakerphone further includes a means 108 for transducing electrical signals into sound waves using, for example, a transducer such as speaker 110. The speaker 110 can be any conventional speaker suitable for transforming electrical signals into audible sound waves.

In accordance with exemplary embodiments, the speakerphone 100 further includes a means 112 for interfacing the sound wave transducing means and the electrical signal transducing means with a line input/output 114 of the speakerphone, the input/output being connected to the two-wire telephone line 102 for transmitting and receiving communication signals. In accordance with the exemplary Figure 1 embodiment, the interfacing means 112 includes multiple signal paths. For example, a signal path is provided between the microphone 106 and the input/output 114. An additional signal path is provided between the input/output 114 and the speaker 110 for delivering signals received by the speakerphone to the speaker. Both signal paths are connected to the line input/output 114 via a conventional four- wire-to-two- wire hybrid 116 of the interfacing means 112.

In accordance with the exemplary Figure 1 embodiment, the signal path between the microphone 106 and the input output 114 includes a microphone amplifier 118, which can be any conventional amplifier suitable for amplification of the electrical signals produced by the microphone 106. Further, because the exemplary Figure 1 interfacing means 112 is used to provide digital signal processing, the signal path between the microphone 106 and the input/output 114 includes an analog-to-digital converter 120. Outputs from the analog-to-digital converter 120 are supplied via a multiplexer

122 and a digital summer 124 to a digital gain control 126. The digital gain control device can, for example, be any conventional digital signal processor implemented function used to control gain within a signal path from the microphone 106 through the input/output 114. Because the input/output 114 of the exemplary Figure 1 embodiment supplies analog signals to the telephone line 102, a digital-to-analog converter 128 is included between the digital gain control device 126 and the hybrid 116.

The signal path between the input/output 114 and the speaker 110 includes an analog summer 130 connected to an output of the hybrid 116. Analog signals received from the telephone line are thus passed through the hybrid 116 and the summer 130 to a programmable gain control device 132. The programmable gain control device 132 can be any conventional device for controlling gain within the signal path between the input/output 114 and the speaker 110. For example, one programmable gain control device which can be used is readily available from Sierra Semiconductor and described in the data sheet for the SCI 1293 A Modem Analog Front End (AFE) device, the contents of which are hereby incorporated by reference in their entirety. The signal output from the programmable gain control device 132 is supplied to the speaker 110 via a speaker amplifier 134.

Due to inclusion of the hybrid 116 in both of the aforementioned signal paths, electrical echos can occur between the microphone 106 and the speaker 110. Further, because the microphone 106 and speaker 110 are in the general vicinity of one another, acoustic echos in the surrounding environment resulting from outputs from speaker 110 can be picked-up by the microphone 106 and create damaging feedback within the speakerphone circuitry. Such feedback can be manifested, for example, as a singing sound output from the speaker 110 or applied over the telephone line 102 to a receiving terminal on the other end (e.g., a receiving speakerphone).

Accordingly, exemplary embodiments of the present invention such as that illustrated in Figure 1 include means 136 for reducing electrical echoing within the interfacing means. In the exemplary Figure 1 embodiment, the electrical echoing reducing means 136 includes an electrical echo canceler for receiving a signal input from the digital gain control device 126. An output from the electrical echo canceler 138 is supplied to the summer 130 via a second digital-to-analog converter 140.

Thus, a compensating signal for providing electrical echo cancellation, representing an estimated echo within the interfacing means, can be supplied to the analog summer 130 to compensate electrical echos in the echo response from hybrid 116 so that microphone signals in the speaker 110 are reduced or eliminated. A feedback loop 142 supplies an output from the summer 130 to the electrical echo canceler 138 via a second analog-to-digital converter 144.

Operation of the electrical echo canceler in response to inputs from the digital gain control device 126 and the feedback loop 142, once coefficients for the electrical echo canceler have been selected, is conventional and need not be described in greater detail herein. However, contrary to conventional use of an electrical echo canceler wherein coefficients are determined after a communication path is established, exemplary embodiments of the present invention are directed to training these coefficients prior to the establishment of the communication path through the telephone line 102. In accordance with exemplary embodiments, such pre-training of the electrical echo canceler can be performed prior to pre-training of a means 146 for reducing acoustic echoing within the interfacing means. As illustrated in the exemplary Figure 1 embodiment, the acoustic echoing reducing means 146 includes a conventional acoustic echo canceler 148 connected to an output of the programmable gain control device 132 via a third analog-to-digital converter 150. An output from the acoustic echo canceler 148 is summed with inputs from microphone 106 to the digital summer 126 to compensate for acoustic feedback received by microphone 106 from the speaker 110. Using the acoustic echoing reducing means, acoustic echos can be reduced or eliminated to improve the quality of the signal supplied by microphone 106 to the input/ output 114 for transmission over the telephone line 102.

Because the cancellation of acoustic echoing may not provide sufficient system margin to permit initial full-duplex operation, conventional systems which use a pre- trained acoustic echo canceler require an initial operation in a half-duplex mode. Alternatively, these conventional systems operate in a full-duplex mode with an annoying singing that substantially degrades the quality of the communications. On the contrary, exemplary embodiments of the present invention are directed to pre- training the electrical echo canceler on the basis that cancellation of electrical echoing can provide a relatively good system margin; thus, upon establishment of communication over the telephone line 102, full-duplex operation can occur immediately.

In accordance with the exemplary Figure 1 embodiment, a pre-training of the electrical echo canceler 138 is provided by a means 152 for adaptively controlling operating characteristics of the electrical echo reducing means 136. The adaptive controlling means 152 includes a controller 154, such as a conventional microprocessor, for selecting the operating characteristics of the electrical echo canceler 138 to balance, or match, an estimated echo of signal paths within the interfacing means to an echo response of the input/output 114 which is connected to the telephone line 102. Coefficients which have been selected to provide the desired operating characteristics can be stored in memory (for example, a random access memory 158) via signal path 164, and can then be later supplied to the electrical echo canceler via signal path 164. The adaptive controlling means 152 further includes means 156 for storing data, such as pre-recorded sound that can be used by the controller 154 during initialization of the operating characteristics in the electrical echo canceler. The data storing means 156 can, for example, be a conventional hard disk drive which is local to the speakerphone 100. The adaptive controlling means 152 can further include a second storage device, such as the random access memory (RAM) 158 for storing operating characteristics, such as the echo canceler coefficients selected in response to a pre-training of the electrical echo canceler.

The adaptive controlling means 152 further includes the multiplexer 122 as a means for inhibiting the sound wave transducing means (that is, the microphone 106 in the exemplary Figure 1 embodiment) during initialization of the echo canceler. In exemplary embodiments, sound can be supplied to the data storing means 156 prior to initialization of the electrical echoing reducing means 136 by directing a digitized voice input from the microphone 106 to the data storing means 156 via path 162. For example, a person can be selected to speak for a predetermined amount of time, such as five seconds, into the microphone 106 during a data storing operation to provide a training signal for the initialization process. During this time, the controller 154 directs the digitized voice outputs from the analog-to-digital converter 120 into the data storing means 156 via path 162. After sufficient data has been recorded, the controller 154 can inhibit the data path from the analog-to-digital converter 120 to the controller 154.

Those skilled in the art will appreciate that the initialization process can be initiated once the dial tone disappears, after the speakerphone is connected to the telephone line 102. An incomplete telephone number can, for example, be dialed into the speakerphone to break any dial tone which may exist over the telephone line 102 without establishing a communication path over the telephone line.

More particularly, once sufficient data has been recorded in the data storing means, the initialization process continues by turning off the microphone and speaker while the speakerphone is taken off hook. To break the dial tone output from the speaker, two dual tone multiple frequency digits can be entered via the speakerphone keypad. The transmission of these two digits is sufficient to break a first dial tone and, if using a private branch exchange, a second dial tone and to connect the speakerphone to the central office. However, transmission of the two digits is typically not sufficient to establish a connection to a second party in accordance with exemplary embodiments. At this point in the initialization process, the impedance of the two- wire telephone line presents an impedance to the speakerphone which is virtually the same as would be present during a communication session (that is, a session where a full seven digit keypad entry is used to establish communication over the two-wire telephone line). During a continuation of the initialization process (that is, subsequent to the connection of the input/output 114 to the telephone line 102, but prior to the establishment of a communication path over the telephone line 102), electrical echos of the pre-recorded sound produced widiin the interfacing means can be reduced. That is, the supply of the initialization data can be supplied from the data storing means via path 166 and multiplexer 122, and operating characteristics (that is, the coefficients) of the electrical echoing reducing means 136 can be adaptively controlled. This initialization process can continue until convergence at the output of summer 130 falls within the predetermined range or limit. This predetermined range or limit can be determined on a case-by-case basis for a given speakerphone and given operating environment.

Once convergence has been achieved, typically within, for example, one second or less, the coefficients of the electrical echo canceler 138 can be stored in the RAM 158 of the adaptive controlling means 152 via path 164. These coefficients will remain suitable provided the speakerphone remains connected to the telephone line 102. However, if the speakerphone is disconnected from the telephone line 102 and reconnected to a different telephone line, an initialization process can be repeated.

Once pre-training of the electrical echo canceler is complete, the telephone can be placed "on hook". Subsequently, the previously established coefficients stored in RAM 158 can be used to set up the electrical echo canceler circuit prior to enabling the microphone when a connection is made. Afterwards, the speakerphone can be immediately operated in a full-duplex mode. During full-duplex operation, either or both of the acoustic echo canceler 148 and electrical echo canceler 138 can be updated during single talk conditions. During double talk conditions (that is, when sound is supplied to the microphone 106 at the same time a communication signal is received by the hybrid 116 via the telephone line 102), coefficients within a acoustic echo canceler 148 and the electrical echo canceler 138 remain frozen.

Those skilled in the art will appreciate that the exemplary embodiment described with respect to Figure 1 is by way of illustration only, and that alternate embodiments will be readily apparent to those skilled in the art. For example, instead of using a predetermined limit for identifying adequate convergence at the output of summer 130 during an initialization process, a predetermined time-out period can also be used. That is, upon expiration of a predetermined time-out period (such as one second), coefficients determined for the electrical echo canceler 138 can be stored in the RAM 158. Further, those skilled in the art will appreciate that all components identified with respect to the exemplary Figure 1 embodiment can be readily obtained off-the- shelf, and assembled in the manner described with respect to Applicants' Figure 1 embodiment. Accordingly, numerous alternate components can be substituted for each of the components illustrated in the Figure 1 embodiment, or various functions described with respect to components of the exemplary Figure 1 embodiment can be combined into single components. However, chip sets suitable for use in providing the acoustic echo cancelling function, the electrical echo cancelling function and other functions associated with the signal paths between the microphone 106 and the input/output 114, and between the input/output 114 and the speaker 110 can be similar to those described in the previously mentioned SCI 1293 A data sheet, the contents of which are hereby incorporated by reference.

Further, those skilled in the art will appreciate that while digital signal processing was performed in the signal path described with respect to the exemplary Figure 1 embodiment, analog processing can also be provided. For example, Figure 2 illustrates an alternative embodiment wherein analog signal processing has been provided for the electrical echo canceler circuit 138. In Figure 2, the analog-to- digital converter 144 and the digital-to-analog converter 140 can be eliminated since the electrical echo canceler 138 is now an analog component. Those skilled in the art will appreciate that because the electrical echo canceler is analog, analog coefficients are used to adjust the estimated echo response output therefrom. Accordingly, in the Figure 2 embodiment, the controller 154 can include analog-to-digital circuitry for converting analog coefficients into digital components that can be stored in the RAM 158, and can further include a digital-to-analog converter for reconverting the stored coefficients into analog values that can be supplied to the electrical echo canceler 138'. It will be appreciated by those skilled in the art that the present invention can be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The presently disclosed embodiments are therefore considered in all respects to be illustrative and not restricted. The scope of the invention is indicated by the appended claims rather than the foregoing description and all changes that come within the meaning and range and equivalence thereof are intended to be embraced therein.

Claims

Claims:

1. Apparatus for establishing a full-duplex signal path comprising: means for transducing sound waves into electrical signals; means for transducing electrical signals into sound waves; means for interfacing said sound wave transducing means and said electrical signal transducing means with an output of the apparatus, said interfacing means further including: means for reducing electrical echoing within said interfacing means; and means for adaptively controlling operating characteristics of said electrical echo reducing means prior to activating said output.

2. Apparatus according to claim 1, wherein said means for adaptively controlling selects said operating characteristic to match a estimated echo within said interfacing means and an echo response from said output.

3. Apparatus according to claim 2, wherein said means for adaptively controlling further includes: means for inhibiting said converting means during an initialization of said electrical echoing reducing means.

4. Apparatus according to claim 3, wherein said means for adaptively controlling further includes: means for storing data for said initialization.

5. Apparatus according to claim 4, wherein said data is pre-recorded sound information.

6. Apparatus according to claim 1, further including: means for reducing acoustic echoing within said interfacing means.

7. Apparatus according to claim 1, wherein said electrical echoing reducing means is formed of digital components.

8. Apparatus according to claim 1, wherein said electrical echoing reducing means is formed of analog components.

9. Apparatus according to claim 1, wherein said output further includes: a two- wire output for connection to a telephone line.

10. A method for establishing a full-duplex signal path comprising the steps: establishing initialization data for electrical echo cancellation between a sound input and a sound output; supplying said initialization data to an electrical echo canceler located between said sound input and said sound output; and reducing electrical echos produced during said step of supplying data by adaptively controlling operating characteristics of said electrical echo canceler.

11. Method according to claim 10, further including the step of: supplying said initialization data to said electrical echo canceler after connecting said sound input and said sound output to a communication path, but prior to establishment of communication between said sound input and said sound output with said communication path.

12. Method according to claim 11, wherein said step of reducing further includes the step of: adaptively controlling said operating characteristics of said electrical echo canceler based on an echo response of said communication path.

13. Method according to claim 12, wherein said step of establishing further includes a step of: storing said initialization data in memory prior to said step of supplying. O 96/14694 PCI7US95/14530

-16-

14. Method according to claim 10, wherein said step of reducing electrical echos further includes a step of: adaptively controlling said operating characteristics of said electrical echo cancellation until a predetermined convergence.

15. Method according to claim 10, wherein said step of reducing electrical echos further includes a step of: adaptively controlling said operating characteristics of said electrical echo canceler until expiration of a predetermined time-out period.