Echo processing devices for single-channel or multichannel communication systems

Claims

The invention claimed is:

1. An echo processing device for attenuating echo components of a direct signal X1n in a return signal Y2n, said device comprising: means for calculating a receive gain Gr.sub.n and a send gain Ge.sub.n; first gain application means for applying the receive gain Gr.sub.n to the direct signal and producing an input signal X2n emitted into an echo generator system; second gain application means for applying the send gain Ge.sub.n to an output signal Y1n from the echo generator system and producing the return signal Y2n; and means for obtaining a coupling variable COR characteristic of the acoustic coupling between the direct signal X1n or the input signal X2n and the output signal Y1n, by calculating a correlation between the direct signal X1n or the input signal X2n and the output signal Y1n; wherein said gain calculation means is configured to calculate the receive gain Gr.sub.n and the send gain Ge.sub.n based on said coupling variable.

2. An echo processing device according to claim 1, comprising means for estimating the instantaneous power of the direct signal X1n or the input signal X2n and the instantaneous power of the output signal Y1n, said gain calculation means being adapted to calculate the receive gain Gr.sub.n and the send gain Ge.sub.n on the basis of a variable G determined as a function of the estimated power of the direct signal or the input signal and the estimated power of the output signal, and as a function of the coupling variable COR, in accordance with the following equation: .times..times..times..times..times..times..times..times..times.- .times..times..times. ##EQU00019## where P1n and P2n are respectively an estimate at the time concerned of the power of the direct signal X1n or the input signal X2n and the power of the output signal Y1n.

3. An echo processing device according to claim 2, in which the gain calculation means determine the receive gain Gr.sub.n and the send gain Ge.sub.n recursively from the following equations: Ge.sub.n=.gamma.Ge.sub.n-1+(1-.gamma.)G Gr.sub.n=1-.delta.Ge.sub.n where Ge.sub.n-1 is the send gain at the preceding calculation time and .gamma.and .delta.are positive constants less than 1.

4. An echo processing device according to claim 1, in which the calculation of the correlation between the direct signal X1n or the input signal X2n and the output signal Y1n is an envelope correlation calculation.

5. An echo processing device according to claim 4, in which, in said envelope correlation calculation, the coupling variable COR is a function of the maximum value Maxcor of the values corr(j) of the correlation between the direct signal X1n or the input signal X2n and the output signal Y1n, said correlation values corr(j) being calculated over a time window considered, and each being obtained from the equation: .function..times..times..times..times..times..times..times..times..times.- .times..times..times..times..times..times. ##EQU00020## in which i is a sampling time in the calculation time window of duration LM, j is a shift value between the input signal X2n and the output signal Y1n , and P1(j) and P2(j) are respectively an estimate of the power of the direct signal X1n or the input signal X2n and an estimate of the power of the output signal Y1n at a time t.

6. An echo processing device according to claim 5, in which the coupling variable COR is linked to the maximum value Maxcor of the correlation values corr(j) calculated over a calculation time window considered from the equation: COR=Exp(k.Maxcor) in which Exp is the exponential function and k is a positive constant.

7. An echo processing device according to claim 1, in which the input signal X2n is emitted into the echo generator system by at least one loudspeaker and the output signal Y1n is obtained from the echo generator system by at least one microphone.

8. An echo processing device according to claim 1, further comprising an echo canceller receiving at its input said input signal X2n emitted into the echo generator system and a signal Y3n from the echo generator system, the echo canceller comprising a finite impulse response identification filter whose response is representative of the response of the echo generator system, and the identification filter being adapted to generate a filtering signal Sn and comprising means for subtracting the filtering signal Sn from the signal Y3n to produce the output signal Y1n that is received at the input of said send gain application means.

9. An echo canceller for attenuating in an output signal Y1n echo components of an input signal X2n emitted into an echo generator system, said device comprising: a finite impulse response identification filter whose response is representative of the response of the echo generator system, receiving the input signal X2n at its input and generating a filtering signal Sn; subtraction means receiving at an input a signal Y3n from the echo generator system, at least one component of which is a response of the echo generator system to the input signal X2n, and the filtering signal Sn, and adapted to subtract the filtering signal Sn from the signal Y3n and to produce the output signal Y1n; means for adapting the coefficients of the identification filter as a function of an adaptation .mu..sub.n; and means for calculating the adaptation .mu..sub.n, said adaptation calculation means comprising means for estimating the power P1n of the input signal X2n and the power P3n of the signal Y3n and means for obtaining a first coupling variable COR2 characteristic of the acoustic coupling between the input signal X2n and the signal Y3n from the echo generator system, by calculating a correlation between the input signal X2n and the signal Y3n; wherein the adaptation .mu..sub.n of the identification filter is calculated as a function of the estimated powers P1n, P3n and as a function of the first coupling variable COR2.

10. A device according to claim 9, in which the adaptation .mu..sub.n is obtained from the equation: .mu..times..times..times..times..alpha..times..times..times..times..times- ..times..times..times..times..times..times. ##EQU00021## in which .alpha.is a positive constant and P1n and P3n are respectively an estimate of the power of the input signal X2n and an estimate of the power of the signal Y3n from the echo generator system at the time concerned.

11. A device according to claim 9, in which the calculation of the correlation between the input signal X2n and the signal Y3n is an envelope correlation calculation.

12. A device according to claim 11, in which the first coupling variable COR2 is a function of the maximum value Maxcor2 of correlation values corr2(j) calculated over a time window considered, each of the correlation values corr2(j) being calculated from the following equation: .function..times..times..times..times..times..times..times..times..times.- .times..times..times..times. ##EQU00022## in which: i is a sampling time in the calculation time window of duration LM and j is a shift value between the input signal X2n and the signal Y3n; and P1(t) and P3(t) are respectively an estimate of the power of the input signal X2n and an estimate of the power of the signal Y3n at the time t concerned.

13. A device according to claim 12, in which the first coupling variable COR2 is linked to the maximum value Maxcor2 of said correlation values corr2(j) by the following equation, in which k is a positive constant: .times..times..times..times. ##EQU00023##

14. An echo canceller according to claim 9, in which the adaptation step calculation means further comprise means for calculating a second coupling variable COR characteristic of the acoustic coupling between the input signal X2n from the echo generator system and the output signal Y1n, the second coupling variable COR being obtained by calculating the correlation between the input signal X2n and the output signal Y1n, and the adaptation step .mu..sub.n of the identification filter being calculated as also a function of the second coupling variable COR.

15. An echo canceller according to claim 14, in which the second coupling variable COR is obtained from an envelope correlation calculation between the input signal X2n and the output signal Y1n.

16. An echo canceller according to claim 15, in which the second coupling variable COR is a function of the maximum value Maxcor of the values corr(j) of the correlation between the input signal X2n and the output signal Y1n, said correlation values corr(j) being calculated over a time window considered and each of them being obtained from the equation: .function..times..times..times..times..times..times..times..times..times.- .times..times..times..times. ##EQU00024## in which i is a sampling time in the calculation window of duration LM, j is a shift value between the input signal X2n and the output signal Y1n, and P1(t) and P2(t) are respectively an estimate of the power of the input signal X2n and an estimate of the power of the output signal Y1n at a time t.

17. An echo canceller according to claim 14, in which the adaptation step .mu..sub.n is calculated from the equation: .mu..times..times..times..times..times..times..alpha..times..times..times- ..times..times..times..times..times..times..times. ##EQU00025## in which .alpha.is a positive constant and P1n and P3n are respectively an estimate of the power of the input signal X2n and an estimate of the power of the signal Y3n from the echo generator system at the time concerned.

18. An echo processing device for a multichannel communications system comprising N receive channels, N being an integer greater than or equal to 2, and M send channels, M being an integer greater than or equal to 1, each of the N receive channels i comprising an output transducer (LSi) that produces a sound pressure wave in response to an input signal X2n(i) derived from a direct signal X1n(i), each of the M send channels j comprising an input transducer (MCj) that converts a sound pressure wave into an output signal Y1n(j), said echo processing device being adapted to attenuate in each output signal Y1n(j) echo components stemming from some or all of the N input signals X2n(i) and resulting from the acoustic coupling between the input transducer of the send channel concerned and some or all of the M output transducers, said device comprising: means for calculating receive gains Gr.sub.n(i) and send gains Ge.sub.n(j); means for applying receive gains Gr.sub.n(i) to each direct signal X1n(i) and producing a corresponding input signal X2n(i); means for applying send gains Ge.sub.n(j) to each output signal Y1n(j) and producing the corresponding return signal Y2n(j); and means for calculating, for each send channel j, N coupling variables COR(j,i), for i varying from 1 to N, each of which being characteristic of the acoustic coupling between the output signal Y1n(j) of the send channel and one of the N input signals X2n(i), each coupling variable COR(j,i) being obtained by calculating a correlation between the corresponding output signal Y1n(j) and the corresponding input signal Y2n(i); wherein said gain calculation means is configured to calculate each receive gain Gr.sub.n(i) and each send gain Ge.sub.n(j) based on the N coupling variables COR(J,i) calculated for the associated send channel j.

19. A device according to claim 18, comprising means for estimating the instantaneous power P1n.sub.i, of each input signal X2n(i) and the instantaneous power P2n.sub.j of each output signal Y1n(j), said send gain calculation means being adapted to calculate each send gain Ge.sub.n(j) on the basis of N variables G(j,i), for i varying from 1 to N, each of which is determined as a function of the estimated power of an input signal X2n(i) and the estimated power of the output signal Y1n(j) of the send channel concerned and as a function of the corresponding coupling variable COR(j,i), each of the variables G(j,i) being obtained from the following equation: .function..times..times..times..times..times..times..times..times..functi- on..times..times..times..times. ##EQU00026## in which P1n .sub.i and P2n .sub.j are respectively an estimate of the power of the input signal X2n(i) concerned and an estimate of the power of the output signal Y1n(j) concerned at the time concerned.

20. A device according to claim 19, in which each send gain Ge.sub.n(j) is determined from the minimum value of the N variables G(j,i), for i varying from 1 to N, calculated for the associated send channel j.

21. A device according to claim 20, in which each send gain Ge.sub.n(j) is determined from the equation: Ge.sub.n(j)=.gamma.Ge.sub.n-1(j)+(1-.gamma.)min.sub.i(G(j,i)) in which Ge.sub.n-1(j) is the send gain of the send channel j at the time of the preceding calculation, .gamma. is a positive constant less than 1, and min.sub.i (G(j,i)) is the minimum value of the N variables G(j,i) for i varying from 1 to N.

22. A device according to claim 21, in which all the receive gains Gr.sub.n(i) have the same value, which is determined from the equation: Gr.sub.n(i)=1-.delta.max.sub.j(Ge.sub.n(j)) in which .gamma. is a positive constant less than 1 and max.sub.j(Ge.sub.n(j)) is the maximum value of the M send gains Ge.sub.n(j), for j varying from 1 to M.

23. A device according to claim 18, in which each of said receive gains Gr.sub.n(i) is equal to 1.

24. A device according to claim 18, in which the calculation of the correlation between an output signal Y1n(j) and an input signal X2n(i) is an envelope correlation calculation.

25. A device according to claim 24, in which, in said envelope correlation calculation, each coupling variable COR(j,i) is a function of the maximum value Maxcor of the values corr.sub.ji(d) of the correlation between the output signal Y1n(j) and the input signal X2n(i), said correlation values corr.sub.ji(d) being calculated over a predefined time window and each obtained from the equation: .function..times..times..times..times..times..times..function..times..tim- es..times..times..function..times..times..times..times..times..times..func- tion. ##EQU00027## in which c is a sampling time in the calculation time window of duration LM d is a shift value between the input signal X2n(i) and the output signal Y1n(j), and P1n.sub.i(t) and P2n.sub.j(t) are respectively an estimate of the power of the input signal X2n(i) and an estimate of the power of the output signal Y1n(j) at a time t.

26. An echo canceller for a multichannel communications system comprising N receive channels, N being an integer greater than or equal to 2, and M send channels, M being an integer greater than or equal to 1, each of the N receive channels i comprising an output transducer (LSi) that produces a sound pressure wave in response to an input signal X2n(i), and each of the M send channels j comprising an input transducer (MCj) that converts a sound pressure wave into an output signal Y1n(j), the echo canceller comprising: for each send channel j, N identification filters Fij with variable coefficients for estimating the acoustic coupling between each of the N output transducers (LSi) and the input transducer (MCj) of the send channel j, and for each filter Fij, means for adapting the coefficients of the filter as a function of an adaptation step .mu..sub.n(i,j) and means for calculating the adaptation step .mu..sub.n(i,j), means for estimating the instantaneous power P1n.sub.i of each input signal X2n(i) and the instantaneous power P2nj of each output signal Y1n(j), and means for calculating, for each send channel j, N coupling variables COR(j,i), for i varying from 1 to N, each of which being characteristic of the acoustic coupling between the output signal Y1n(j) of the send channel j and one of the N input signals X2n (i), each coupling variable COR(j,i) being obtained by calculating a correlation between the output signal Y1n(j) and the input signal X2n(i); wherein the means for calculating the adaptation step .mu..sub.n(i,j) for a filter Fij associated with a receive channel i and a send channel j is configured to calculate the adaptation step .mu..sub.n(i,j) as a function of the powers P1n.sub.i, for i varying from 1 to N, estimated for the N receive channels, as a function of the estimated power P2nj of the send channel j, and as a function of the N coupling variables COR(j,i), for i varying from 1 to N, associated with the send channel j.

27. A device according to claim 26, in which an adaptation step .mu..sub.n(i,j) for a filter Fij associated with a receive channel i and a send channel j is obtained from the following equation, in which b.sub.i is a positive constant: .mu..function..times..times..times..times..times..times..times..times..fu- nction..times..times..times..times..noteq..times..times..function..times..- times..times..times. ##EQU00028##

28. A device according to claim 26, in which the calculation of the correlation between the output signal Y1n(j) and the input signal X2n (i) is an envelope correlation calculation.

29. A device according to claim 28, in which the coupling variable COR(j,i) is a function of the maximum value Maxcor(j,i) of the correlation values corr.sub.ji(d), calculated over a time window considered, each of the correlation values corr.sub.ji(d) being calculated from the equation: .function..times..times..times..times..times..times..function..times..tim- es..times..times..function..times..times..times..times..times..times..func- tion. ##EQU00029## in which c is a sampling time in the calculation time window of duration LM, d is a shift value between the input signal X2n(i) and the output signal Y1n(i), and P1n.sub.j(t) and P2n .sub.j(t) are respectively an estimate of the power of the input signal X2n(i) and an estimate of the power of the output signal Y1n(j) at a time t.

30. A device according to claim 29, in which the coupling variable COR(j,i) is linked to the maximum value Maxcor(j,i) of said correlation values corr.sub.ji(d) by the following equation, in which k is a positive constant: .function..function. ##EQU00030##

31. A device according to claim 26, in which each filter Fij associated with a receive channel i, and a send channel j generates a filtering signal that is subtracted from the output signal Y1n(j) to provide a filtered signal Y2n(j), said device further comprising means for calculating, for each send channel j, N second coupling variables COR2(j,i), for i varying from 1 to N, each of which being characteristic of the acoustic coupling between the filtered signal X2n(j) from the send channel and one of the N input signals X2n(i), the adaptation step .mu..sub.n(i,j) of an identification filter Fij associated with a receive channel i and a send channel j being calculated also as a function of said N second coupling variables COR2(j,i).

32. A device according to claim 31, in which an adaptation step .mu..sub.n(i,j) for a filter Fij associated with a receive channel i and a send channel j is obtained from the following equation, in which b.sub.i is a positive constant: .mu..function..function..function..times..times..function..times..times..- noteq..times..times..function. ##EQU00031##

33. A device according to claim 31, further comprising, for each pair comprising a receive channel i and a send channel j, gain application means for applying a receive gain Gr.sub.n(i) to the input signal X2n(i) and a send gain Ge.sub.n(j) to the filtered signal Y2n(j), said gains Gr.sub.n(i), Ge.sub.n(j) being calculated on the basis of the N second coupling variables COR2(j,i) determined for the send channel i.

34. An echo processing device for attenuating echo components of a direct signal X1nin a return signal X2n, said device comprising: means for calculating a receive gain Gr.sub.n and a send gain Ge.sub.n; first gain application means for applying the receive gain Gr.sub.n to the direct signal and producing an input signal X2n emitted into an echo generator system; second gain application means for applying the send gain Ge.sub.n to an output signal Y1n from the echo generator system and producing the return signal Y2n; means for obtaining a first coupling variable COR characteristic of the acoustic coupling between the direct signal X1n or the input signal X2n and the output signal Y1n, by calculating the correlation between the direct signal X1n or the input signal X2n and the output signal Y1n, said gain calculation means being adapted to calculate the receive gain Gr.sub.n and the send gain Ge.sub.n based on said first coupling variable COR; and an echo canceller receiving at its input said input signal X2n emitted into the echo generator system and a signal Y3n from the echo generator system, the echo canceller comprising: a finite impulse response identification filter whose response is representative of the response of the echo generator system, the identification filter being adapted to generate a filtering signal Sn and comprising means for subtracting the filtering signal Sn from the signal Y3n to produce said output signal Y1n that is received at the input of said second gain application means, means for adapting the coefficients of the identification filter as a function of an adaptation step .mu..sub.n; and means for calculating the adaptation step .mu..sub.n, said adaptation step calculation means comprising means for estimating the power P1n of the input signal X2n or the direct signal X1n and the power P3n of the signal Y3n, and means for obtaining a second coupling variable COR2 characteristic of the acoustic coupling between the input signal X2n or the direct signal X1n, and the signal Y3n from the echo generator system, by calculating a correlation between the input signal X2n or the direct signal X1n, and the signal Y3n; wherein the adaptation .mu..sub.n of the identification filter is calculated as a function of the estimated powers P1n, P3n and as a function of said second coupling variable COR2.

35. An echo processing device according to claim 34, in which said adaptation step .mu..sub.n of the identification filter is calculated also as a function of the first coupling variable COR.

RELATED APPLICATIONS

This is a U.S. National Phase Application under 35 USC 371 of International Application PCT/FR2003/001874, filed on 18 Jun. 2003.

FIELD OF THE INVENTION

The field of the present invention is that of communications. The invention relates more particularly to variable-gain and/or adaptive filtering acoustic echo processing devices for attenuating echo components of a direct signal in a return signal. The invention applies to single-channel and multichannel communications systems.

BACKGROUND OF THE INVENTION

Acoustic echoes occur primarily in certain types of communication in which a remote user terminal comprises one or more directional microphones and one or more loudspeakers instead of an earpiece. Examples include audioconference equipment and hands-free telephones, such as mobile telephones. The source of the echoes is simple: failing special precautions, sound emitted by the loudspeaker(s) is reflected many times (from walls, the ceiling, etc.), constituting as many different echoes which are picked up by the microphone(s) on the same terms as wanted speech. Thus the combination of the loudspeaker(s), the microphone(s), and their physical environment constitutes an echo generator system.

The acoustic echo problem has been the subject of much research, both in the case of single-channel systems (one microphone and one loudspeaker) and in the case of multichannel systems (a plurality of microphones and a plurality of loudspeakers). The echo problem in the multichannel situation is similar to that in the single-channel situation except that all possible acoustic couplings between the various microphones and loudspeakers must be considered.

The echo processing techniques most widely used include echo suppression techniques using gain variation and echo cancellation techniques using adaptive filtering.

In a variable-gain echo suppression system, a receive gain is applied to the signal for application to the loudspeaker (the direct signal at the input of the echo generator system) and a send gain is applied to the signal coming from the microphone (at the output of the echo generator system), forming the return signal. An echo suppression system of this type is described in French Patent No. 2 748 184.

Receive voice activity detectors (RVAD), send voice activity detectors (SVAD), and double speech detectors (DSD) typically supply the necessary information to the modules that calculate the send and receive gains. Thus when the remote party is speaking (detected by the RVAD), the send gain is reduced to attenuate the echo. If the local party begins to speak (detected by the SVAD), this constraint on the send gain is removed and the receive gain is reduced. In the event of double speech (both parties speaking simultaneously, detected by the DSD), either a comparator determines which speaker is louder and gives priority to that speaker's sending direction or an intermediate setting of the send and receive gains is established.

In an acoustic echo canceller (AEC) using adaptive filtering, an identification filter estimates the acoustic coupling between the loudspeaker and the microphone and generates a signal that is used to cancel the echo. The identification filter is conventionally a programmable finite impulse response filter whose coefficients need to be adapted by a predetermined algorithm for updating coefficients using an adaptation step. The coefficients are adapted on the basis of the signal to be applied to the loudspeaker. An echo canceller of this type is described in French Patent No. 2 738 695.

A variable gain echo suppression system is often combined with an echo canceller to eliminate the residual echo that remains after echo cancellation.

However, the above-mentioned echo processing systems have the drawback that they are not able to take account of variations in the acoustic coupling between the loudspeaker and the microphone if those variations are independent of the signal applied to the loudspeaker.

This is the case, for example, if there is an external facility for adjusting the sound level reproduced by the loudspeaker (for example by means of a potentiometer). Any variation in the reproduced sound level modifies the acoustic coupling between the loudspeaker and the microphone and therefore the echo(es) picked up by the microphone. The echo processing system takes account only of the signal that is applied to the loudspeaker, and not of the sound that is actually reproduced by the loudspeaker, and is therefore unable to take this kind of modification of the acoustic coupling into account in its calculation process.

For example, if the sound reproduction level is reduced after the system has been initialized with a maximum sound level setting, in a double speech situation the remote speech emitted by the loudspeaker may be broken up or truncated.

Similarly, if the microphone and the loudspeaker in the communications terminal being used are physically independent of each other, the distance between them may be varied, which varies the acoustic coupling between the loudspeaker and the microphone, with the same consequences.

The problem is the same in a multichannel situation except that it generalized to the multiple couplings between the various microphones and loudspeakers.

SUMMARY OF THE INVENTION

One particular object of the present invention is to remedy the drawbacks of prior art echo processing systems described hereinabove.

To this end, in a first aspect, the present invention provides an echo processing device for attenuating echo components of a direct signal X1n in a return signal Y2n, said device comprising: means for calculating a receive gain Gr.sub.n and a send gain Ge.sub.n; first gain application means for applying the receive gain Gr.sub.n to the direct signal and producing an input signal X2n emitted into an echo generator system; and second gain application means for applying the send gain Ge.sub.n to an output signal Y1n from the echo generator system and producing the return signal Y2n.

According to an embodiment of the invention, this echo processing device is noteworthy in that it further comprises means for calculating a coupling variable COR characteristic of the acoustic coupling between the direct signal X1n or the input signal X2n and the output signal Y1n and in that said gain calculation means are adapted to calculate the receive gain Gr.sub.n and the send gain Ge.sub.n on the basis of said coupling variable.

Taking account in the device of the real acoustic coupling between the loudspeaker and the microphone when controlling the variation of the receive and/or send gain applied automatically adapts the sound quality of the sent signal and the received signal as a function of changes in the acoustic environment of the echo processing device and the relative position of the transducers (loudspeaker(s), microphone(s)) and as a function of the sound reproduction level chosen by the user, for example.

According to one particular feature of the invention, the echo processing device comprises means for estimating the instantaneous power of the direct signal X1n or the input signal X2n and the instantaneous power of the output signal Y1n. The gain calculation means are adapted to calculate the receive gain Gr.sub.n and the send gain Ge.sub.n on the basis of a variable G determined as a function of the estimated power of the direct signal or the input signal and the estimated power of the output signal and as a function of the coupling variable COR, in accordance with the following equation:

.times..times..times..times..times..times..times..times..times..times..tim- es..times. ##EQU00001##

where P1n and P2n are respectively an estimate of the power of the direct signal X1n or the input signal X2n and an estimate of the power of the output signal Y1n at the time concerned.

The term "CORP1n" in the expression for the variable G represents the energy of the sound actually picked up by the microphone, and therefore taking into account all external adjustments that are not "seen" by the system (for example the sound reproduction level). The variable G therefore varies automatically as a function of real changes in loudspeaker/microphone acoustic coupling and the send and receive gains are therefore adapted automatically.

In a second aspect, the invention provides an echo canceller for attenuating, in an output signal Y1.sub.n, echo components of an input signal X2n emitted into an echo generator system, said device comprising: a finite impulse response identification filter whose response is representative of the response of the echo generator system, receiving the input signal X2n at its input and generating a filtering signal Sn; subtraction means receiving at an input a signal Y3n from the echo generator system, at least one component of which is a response of the echo generator system to the input signal X2n, and the filtering signal Sn, and adapted to subtract the filtering signal Sn from the signal Y3n and to produce the output signal Y1n; means for adapting the coefficients of the identification filter as a function of an adaptation step .mu..sub.n; and means for calculating the adaptation step .mu..sub.n.

This echo canceller is noteworthy in that the adaptation step calculation means comprise means for estimating the power P1n of the input signal X2n and the power P3n of the signal Y3n and means for calculating a first coupling variable COR2 characteristic of the acoustic coupling between the input signal X2n and the signal Y3n from the echo generator system, the adaptation step .mu..sub.n of the identification filter being calculated as a function of the estimated powers P1n, P3n and as a function of the first coupling variable COR2.

Evaluating the above coupling variable COR2 means that the adaptation step of the filter may be "driven" as a function of the real acoustic coupling between the input signal and the output signal of the echo generator system. This improves the responsiveness of the echo canceller as a function of changes in the acoustic environment of the device, and therefore improves the result of echo processing.

In a preferred embodiment, the adaptation step .mu..sub.n is obtained from the equation:

.mu..times..times..times..times..alpha..times..times..times..times..times.- .times..times..times..times..times..times. ##EQU00002##

in which .alpha.is a positive constant and P1n and P3n are respectively an estimate of the power of the input signal X2n and an estimate of the power of the signal Y3n from the echo generator system, at the time concerned.

In one embodiment, the adaptation step calculation means further comprise means for calculating a second coupling variable COR characteristic of the acoustic coupling between the input signal X2n from the echo generator system and the output signal Y1n, the second coupling variable COR being obtained by calculating the correlation between the input signal X2n and the output signal Y1n, and the adaptation step .mu..sub.n of the identification filter being calculated as a function of the second coupling variable COR.

By additionally taking account of the second coupling variable COR, it is possible to determine the state of convergence of the identification filter and thus to apply finer control of the adaptation step.

In a third aspect, the invention provides an echo processing device for a multichannel communications system comprising N receive channels, N being an integer greater than or equal to 2, and M send channels, M being an integer greater than or equal to 1, each of the N receive channels i comprising an output transducer that produces a sound pressure wave in response to an input signal X2n(i) derived from a direct signal X1n(i), each of the M send channels j comprising an input transducer that converts a sound pressure wave into an output signal Y1n(j), and said echo processing device being adapted to attenuate, in each output signal Y1n(j), echo components stemming from some or all of the N input signals X2n(i) and resulting from the acoustic coupling between the input transducer of the send channel concerned and some or all of the M output transducers.

According to an embodiment of the invention the device is noteworthy in that it comprises: means for calculating receive gains Gr.sub.n(i) and send gains Ge.sub.n(j); means for applying a receive gain Gr.sub.n(i) to each direct signal X1n(i) and producing the corresponding input signal X2n(i): means for applying a send gain Ge.sub.n(j) to each output signal Y1n(j) and producing the corresponding return signal Y2n(j); and means for calculating, for each send channel j, N coupling variables COR(j,i), for i varying from 1 to N, each of which is characteristic of the acoustic coupling between the output signal Y1n(j) of the send channel and one of the N input signals X2n(i);

said gain calculation means being adapted to calculate each receive gain Grn(i) and each send gain Ge.sub.n(j) on the basis of the N coupling variables COR(j,i) calculated for the associated send channel j.

The advantages of this mode of calculating gains in respect of a given pair of send and receive channels (i, j) are of the same kind as are obtained with a variable gain single-channel device of the invention, as briefly set out hereinabove.

In a preferred embodiment of the invention, the echo processing device comprises means for estimating the instantaneous power P1n.sub.i of each input signal X2n(i) and the instantaneous power P2n.sub.j of each output signal Y1n(j), said send gain calculation means being adapted to calculate each send gain Gen(j) on the basis of N variables G(j,i), for i varying from 1 to N, each of which is determined as a function of the estimated power of an input signal X2n(i) and the estimated power of the output signal Y1n(j) of the send channel concerned and as a function of the corresponding coupling variable COR(j,i), and each of the variables G(j,i) being obtained from the following equation:

.function..times..times..times..times..times..times..times..times..functio- n..times..times..times..times. ##EQU00003##

in which P1n.sub.i and P2nj are respectively an estimate of the power of the input signal X2n(i) concerned and of the power of the output signal Y1n(j) concerned at the time concerned.

In a fourth aspect, the invention provides an echo canceller for a multichannel communications system comprising N receive channels, N being an integer greater than or equal to 2, and M send channels, M being an integer greater than or equal to 1, each of the N receive channels i comprising an output transducer that produces a sound pressure wave in response to an input signal X2n(i), and each of the M send channels j comprising an input transducer that converts a sound pressure wave into an output signal Y1n(j), the device comprising: for each send channel j, N identification filters Fij with variable coefficients for estimating the acoustic coupling between each of the N output transducers and the input transducer of the send channel j, and for each filter Fij, means for adapting the coefficients of the filter as a function of an adaptation step .mu..sub.n(i,j) and means for calculating the adaptation step .mu..sub.n(i,j).

According to an embodiment of the invention, this device is noteworthy in that it comprises: means for estimating the instantaneous power P1n.sub.i of each input signal X2n(i) and the instantaneous power P2n.sub.j of each output signal Y1n(j), and means for calculating, for each send channel j, N coupling variables COR(j,i), for i varying from 1 to N, each of which being characteristic of the acoustic coupling between the output signal Y1n(j) of the same channel and one of the N input signals X2n(i), the means for calculating the adaptation step .mu..sub.n(i,j) for a filter Fij associated with a receive channel i and a send channel j, being adapted to calculate the adaptation step .mu..sub.n(i,j) as a function of the powers P1n.sub.i, for i varying from 1 to N, estimated for the N receive channels, as a function of the power P2n.sub.j estimated for the send channel j, and as a function of the N coupling variables COR(j,i), for i varying from 1 to N, associated with the send channel j.

In a preferred embodiment, an adaptation step .mu..sub.n(i,j) for a filter Fij associated with a receive channel i and a send channel j is obtained from the following equation, in which b.sub.i is a positive constant:

.mu..function..times..times..times..times..times..times..times..times..fun- ction..times..times..times..times..noteq..times..times..function..times..t- imes..times..times. ##EQU00004##

Further features and advantages of the invention will become apparent in the course of the following description of preferred embodiments of the invention, which is given with reference to the appended drawings, in which:

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram of a variable gain single-channel echo processing device according to a first embodiment of the invention;

FIG. 2 is a block diagram of a single-channel echo processing device combining a variable gain system and an echo canceller, according to a second embodiment of the invention;

FIG. 3 is a block diagram of a single-channel echo canceller according to a third embodiment of the invention;

FIG. 4 is a block diagram of a single-channel echo canceller according to a fourth embodiment of the invention;

FIG. 5 is a block diagram of a single-channel echo processing device of the invention combining the features of the first and fourth embodiments of the invention;

FIG. 6 is a block diagram of a variable-gain multichannel echo processing device according to a fifth embodiment of the invention; and

FIG. 7 is a block diagram of a multichannel echo canceller according to a sixth embodiment of the invention.

DETAILED DESCRIPTION OF THE DRAWINGS

FIG. 1 shows a variable-gain single-channel echo processing device according to a first embodiment of the invention. This device is integrated into a hands-free telephone, for example.

As shown in FIG. 1, the device receives and sends digital signals X1n, Y2n respectively called the direct signal and the return signal.

The echo processing device comprises a module 36 for calculating the receive gain (Gr.sub.n) and the send gain (Ge.sub.n). The receive gain Gr.sub.n is applied to the direct signal X1n by a multiplier 10 to obtain an input signal X2n that is emitted into an echo generator system 26.

Similarly, the send gain Ge.sub.n is applied to an output signal Y1n from the echo generator system by a multiplier 12 to produce the return signal Y2n.

The input signal X2n is delivered to a loudspeaker 22 via a digital-to-analog converter (DAC) 14 and an amplifier 18. The amplifier 18 is typically a variable-gain amplifier so that a user of the device may adjust the volume of the sound reproduced by the loudspeaker 22 to suit his convenience.

In a similar manner, the output signal Y1n is obtained from a microphone 24 via an amplifier 20 and an analog-to-digital converter (ADC) 16.

In the embodiment shown, the device comprises a single loudspeaker 22 and a single microphone 24 forming part of the echo generator system 26. However, the device of the invention shown in FIG. 1 may equally well be applied to a system in which the input signal X2n is emitted into the echo generator system by a plurality of loudspeakers 22 reproducing the same sound signal and the output signal Yin is obtained from the echo generator system by means of a plurality of microphones 24.

According to the invention, the echo processing device comprises a module 30 for calculating a coupling variable COR characteristic of the acoustic coupling between the direct signal X1n or the input signal X2n and the output signal Y1n.

To this end, the calculation module 30 comprises a calculation unit 34. The coupling variable COR is calculated by the unit 34 and then used by the gain calculation module 36 to calculate the receive gain Gr.sub.n and the send gain Ge.sub.n.

In the embodiment shown in FIG. 1, the module 30 for calculating the coupling variable COR comprises a unit 28 for estimating the instantaneous power P1n of the input signal X2n and/or the direct signal X1n and an unit for estimating the instantaneous power P2n of the output signal Y1n.

In this embodiment, the gain calculation module 36 is designed to calculate the receive gain Gr.sub.n and the send gain Ge.sub.n on the basis of a variable G calculated by the calculation unit 34 as a function firstly of the estimated power P1n of the direct signal and/or the input signal and the estimated power P2n of the output signal, and secondly as a function of the coupling variable COR.

In a preferred embodiment of the invention, the variable G is determined by the calculation unit 34 from the following equation:

.times..times..times..times..times..times..times..times..times..times..tim- es..times. ##EQU00005##

where P1n and P2n are respectively an estimate of the power of the direct signal X1n or the input signal X2n and an estimate of the power of the output signal Y1n, at the time concerned.

Accordingly, strong coupling (i.e. a high level of correlation) between the direct signal X1