JP2002368658A - Multi-channel echo cancellation apparatus, method, recording medium, and audio communication system - Google Patents

Abstract

(57) [Problem] To provide a multi-channel echo canceller, a method, a recording medium, and a voice communication system capable of estimating an echo path without deteriorating the sound quality of the voice of a received signal and canceling the echo. A first speech detection unit 106 for detecting a speaker 103 _i plurality of channels, the audio from the received signal
_i , a second voice detection unit 104 for detecting voice from the microphone input signal, and a non-correlation signal for adding uncorrelated signals between channels to the received signal of each channel when the received signal and the microphone input signal have no voice. Correlation signal adding section 102
And an adaptive filter 101 for estimating an echo path between the speaker and the microphone using the received signal to which the uncorrelated signal is added, and a subtractor 105 for subtracting a pseudo echo from the microphone input signal.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multi-channel echo canceling apparatus using a plurality of speakers and microphones, a method, a recording medium, and a voice communication system.

[0002]

2. Description of the Related Art Conventionally, this type of multi-channel echo canceller includes an adaptive filter 101, an uncorrelated signal adding unit 102, speakers 103 _{1 to} 103 _m , a microphone 104, and a subtracting unit 105, as shown in FIG. The uncorrelated signal adding unit 102 adds independent white noise or the like (uncorrelated signal) having no correlation between channels to the received signal of each channel, and the received signal to which the uncorrelated signal is added is output to each speaker. 103 _{1 to} 103 _m , the echo signal is input to the microphone 104 as an echo signal, and the adaptive filter 101 estimates an echo path between a plurality of speakers-microphones from the received signal to which the uncorrelated signal is added and the microphone input signal. , A pseudo echo (adaptive filter 10) from the microphone input signal by the subtractor 105.
(Suzuki, Sakauchi, Shimauchi, Haneda, "Study of preprocessing method for improving convergence in stereo echo canceller", Sound Lecture Book 3-5-10, March 1998).

In another multi-channel echo canceling method, a signal obtained by squaring a received signal and adding a gain thereto is added to a corresponding received signal to obtain a cross-correlation variation-added signal. By performing echo cancellation processing using the correlation variation added signal, the cross-correlation between the received signals is actively changed so as not to cause any auditory inconvenience, and an attempt has been made to correctly estimate the echo path (JP-A-11-2519).
No. 74).

In another multi-channel echo canceling method, a low cross-correlation signal (additional signal) is added to a reception signal to drive a speaker, and each reception signal and the additional signal are input to different adaptive filters. The difference between the sum of the outputs of all the adaptive filters and the echo signal is regarded as an error, the coefficients of the adaptive filters are updated so as to reduce this error, and the sum of the outputs of the respective adaptive filters is regarded as a pseudo echo. Is used as the echo cancellation signal, and the reception signal and the additional signal are separately processed (Japanese Patent No. 3073976).
No.).

[0005]

However, in such a conventional multi-channel echo canceling apparatus and method, an uncorrelated signal is also added to the voice section of the received signal, so that the voice of the received signal reproduced from the speaker is reproduced. However, there is a problem that the sound quality is deteriorated.

SUMMARY OF THE INVENTION The present invention has been made to solve such a problem, and estimates an echo path between a speaker and a microphone without deteriorating the sound quality of a received signal.
Multi-channel echo cancellation apparatus and method capable of canceling echo
A recording medium and an audio communication system are provided.

[0007]

SUMMARY OF THE INVENTION A multi-channel echo canceller according to the present invention detects as many speakers as the number of channels for reproducing the received signal of each channel for each channel, and detects sound for each channel from the received signal of each channel. First for the number of channels
And a second sound detection unit for detecting sound from the microphone input signal, and sound detection by the first sound detection unit and the second sound detection unit, thereby converting the reception signal and the microphone input signal into sound. When there is no signal, an uncorrelated signal adding unit that adds a signal that is uncorrelated between channels to a received signal of each channel, and using the received signal to which the uncorrelated signal is added, the number of channels between the speaker and the microphone is reduced. And a subtraction unit that subtracts a pseudo echo output from the adaptive filter from the microphone input signal. With this configuration, the uncorrelated signal is not added to the voice section of the received signal, so that the echo path between the speaker and the microphone can be estimated and the echo can be eliminated without deteriorating the sound quality of the voice of the received signal.

Further, the multi-channel echo canceling apparatus of the present invention has speakers for the number of channels for reproducing the received signal of each channel for each channel, and speakers for the number of channels for dividing the received signal of each channel into a plurality of frequency bands. A first band dividing unit, a first sound detecting unit for detecting the sound of an arbitrary band of an arbitrary channel from a received signal of each band-divided channel, and a number of first sound detecting units equal to the number of channels multiplied by the number of band divisions; A second band division unit for dividing the input signal into bands, a second sound detection unit for the number of band divisions for detecting the sound of the band from the band-divided microphone input signal, A third band division unit for dividing the reception signal of the channel into a plurality of frequency bands for the number of channels;
When there is no sound in the received signal and the microphone input signal by the sound in the arbitrary band by the sound detection unit and the second sound detection unit, the signal uncorrelated between the channels in the reception signal of each channel in the band. Uncorrelated signal adding units for the number of band divisions to be added, subband adaptive filters for estimating echo paths for the number of channels between the loudspeaker and the microphone using the uncorrelated signals, and the subband from the subband microphone input signal. A subband subtraction unit for reducing a subband pseudo echo output from the adaptive filter,
And a band synthesizing unit for synthesizing the band-divided sub-band signals after echo cancellation. With this configuration, since the uncorrelated signal is not added to the voice section of the received signal, the echo path between the speaker and the microphone can be estimated for each subband and the echo can be eliminated without deteriorating the sound quality of the voice of the received signal. Becomes

According to the multi-channel echo canceling method of the present invention, a signal reproducing procedure for reproducing the received signals of each channel corresponding to the number of channels for each channel, and detecting a sound for each channel from the received signals of each channel corresponding to the number of channels. A first voice detection procedure, a second voice detection procedure for detecting voice from a microphone input signal, the first voice detection procedure, and the second voice detection procedure.
By the voice detection according to the voice detection procedure, when there is no voice in the reception signal and the microphone input signal, a non-correlation signal addition procedure of adding a non-correlation signal between channels to the reception signal of each channel; It has an echo path estimation procedure for estimating the number of echo paths for the number of channels between the speaker and the microphone using the added received signal, and a subtraction procedure for subtracting the pseudo echo output in the path estimation procedure from the microphone input signal. . In this way,
Since the uncorrelated signal is not added to the voice section of the received signal, the echo path between the speaker and the microphone can be estimated and the echo can be eliminated without deteriorating the sound quality of the voice of the received signal.

[0010] The recording medium of the present invention provides a computer with a signal reproducing procedure for reproducing the received signals of each channel for each channel, and detecting the sound for each channel from the received signals of each channel for the number of channels. A first voice detection procedure, a second voice detection procedure for detecting voice from a microphone input signal, the first voice detection procedure, and the second voice detection procedure.
By the voice detection according to the voice detection procedure, when there is no voice in the reception signal and the microphone input signal, a non-correlation signal addition procedure of adding a non-correlation signal between channels to the reception signal of each channel; In order to execute an echo path estimation procedure for estimating the number of echo paths for the number of channels between the speaker and the microphone using the added received signal, and an echo cancellation procedure for subtracting a pseudo echo output from the microphone input signal in the echo path estimation procedure. Which is readable by the computer. With this recording medium, the uncorrelated signal is not added to the voice section of the received signal, so that the echo path between the speaker and the microphone can be estimated and the echo can be eliminated without deteriorating the sound quality of the voice of the received signal. .

The voice communication system according to the present invention has a configuration provided with the multi-channel echo canceling device and control means for controlling the multi-channel echo canceling device by a program recorded on the recording medium. With this configuration, since the uncorrelated signal is not added to the voice section of the received signal, the echo path between the speaker and the microphone is estimated without deteriorating the sound quality of the voice of the received signal.
The echo can be eliminated.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described.
This will be described with reference to the drawings. [First Embodiment] As shown in FIG.
The multi-channel echo canceling apparatus of the embodiment
Of the number of channels for reproducing the received
PEAKA 103 ₁~ 103_mAnd from the received signal of each channel
First sound for the number of channels for which sound is detected for each channel
Detects sound from detection unit 106 and microphone input signal
A second voice detection unit 107 that performs
6 and the voice detection by the second voice detection unit 107,
When there is no sound in the received signal and microphone input signal
In addition, the received signal of each channel has a signal that is uncorrelated between channels.
And a non-correlated signal adding section 102 for adding
Speaker 103 using the reception signal with₁-10
3_mEcho pads for the number of channels between the
Filter 101 and microphone input
A pseudo echo output from the adaptive filter 101 is obtained from the signal.
And a subtraction unit 105 for subtraction.

First, the configuration will be described. As shown in FIG. 1, the first voice detection unit 106 includes voice detection units 106 _{1 to} 106 _{m for the} number of channels. Here, the processing of the first voice detection unit 106 _i will be described with reference to FIG.
i indicates a channel number, and i = 1 to m. As shown in FIG. 2, the first voice detection unit 106 _i outputs the received signal x
One frame of _i (k) (frame length: 10 to 100 [ms]
), And the power x _i _pow (k) of the received signal is calculated for each frame (step S201), and the first noise power estimation value nx _i _pow is calculated.
(K-1) (step S202). As a result of this comparison, if x _{i —} pow (k) ≦ nx _{i —} pow (k−1) (yes in step S202), x _{i —} pow
Let (k) be the first estimated noise power value (step S2)
03). Also, x _i _pow (k)> nx _i _pow (k−
In the case of 1) (no in step S202), the first noise power estimation value nx _i _pow (k−1) is multiplied by the first update amount Lup _x (> 1.0), and nx _i _pow (k) ) Is updated (step S204).

The aforementioned steps S203 and S2
04, the received signal power x _i _pow of the current frame
(K) is the first noise power estimated value nx _i _pow (k)
Is multiplied by a first voice detection multiplier α _x (> 1.0) to determine whether or not it is equal to or higher than a voice detection threshold (step S2).
05). As a result of this determination, x _{i —} pow (k) ≧ nx _{i —} p
ow (k) × α _x , there is a voice (step S20).
6) is determined. Also, as a result of the determination, x _i _pow (k)
If <nx _{i —} pow (k) × α _x , it is determined that there is no voice (step S207).

Next, the processing of the second voice detection unit 107 will be described with reference to FIG. As shown in FIG.
Of the microphone input signal y (k)
Is buffered, and the received signal power y_pow (k) is calculated for each frame (step S301), and the second noise power estimated value ny_ is calculated.
pow (k-1) (step S302). As a result of this comparison, y_pow (k) ≦ ny_pow (k−
In the case of 1), y_pow (k) is used as the second noise power estimation value (step S303). Also, y_pow
(K)> ny_pow (k-1) (Step S3
02), the second noise power estimated value ny_po
Multiply w (k-1) by the second update amount Lup _y (> 1.0) to update ny_pow (k) (step S30).
4).

The aforementioned steps S303 and S3
04, the received signal power y_pow of the current frame
It is determined whether or not (k) is equal to or greater than a sound detection threshold value calculated by multiplying the second estimated noise power value ny_pow (k) by the second sound detection multiplier α _y (> 1.0) (step S30).
5). As a result of this determination, y_pow (k) ≧ ny_pow
If (k) × α _y , there is a voice (step S306)
Is determined. Also, y_pow (k) <ny_pow
In the case of (k) × α _y (No in step S305),
It is determined that there is no voice (step S307).

When the first voice detector 106 _i and the second voice detector 107 both determine that there is no voice, the decorrelation signal adding unit 102 calculates each of the signals as shown in the following equation (1). The received signal x _i (k) of the channel is, for example, a signal n _i (k) uncorrelated between channels such as white noise.
Is added.

(Equation 1) Note that the uncorrelated signal can be added in a stepwise manner so as to alleviate the sense of incongruity in hearing. Further, as shown in the following equation (2), the level of the received signal x _i (k) is reduced in the decorrelation signal adding section 102,
It is possible to prevent the overall level from rising due to the addition of the uncorrelated signal.

(Equation 2)

The adaptive filter 101 is used when the first voice detector 106 _i and the second voice detector 107 both determine that there is no voice, that is, when the uncorrelated signal is added to the received signal. Only, using a received signal of all channels to which an uncorrelated signal is added and an error signal of all channels, a learning identification method, an affine projection algorithm,
(F) The echo path is estimated by an adaptive algorithm such as the RLS algorithm, and the pseudo echo is calculated by convolving the filter coefficient with the received signal.

Further, the subtraction unit 105 is provided with the adaptive filter 10.
The error signal after echo cancellation is calculated by subtracting the pseudo echo, which is the output of 1, from the microphone input signal.

Next, the operation principle of this embodiment will be described with reference to FIG. First, the first sound detector 106 _i, detects whether there is audio in the received signal (step S401).

Next, the second voice detector 107 detects whether or not there is voice in the microphone input signal (step S402).

[0022] Then, in the decorrelated signal addition section 102, the first sound detector 106 _i and the second voice detection section 1
If no voice is detected in step 07, a signal uncorrelated between channels or an independent signal is added to the received signal (step S403).

[0023] Then, in the adaptive filter 101, if the speech is not detected by the first speech detection unit 106 _i and the second voice detection unit 107, a reception signal uncorrelated signal is added, the microphone input An echo path is estimated from the signal and a pseudo echo is calculated (step S404).

Next, the subtractor 105 eliminates the echo by subtracting the pseudo echo output from the adaptive filter from the microphone input signal (step S).
405).

As described above, the first embodiment of the present invention
The multi-channel echo canceller according to
Speakers 1 for the number of channels for reproducing signals for each channel
03 ₁~ 103_mAnd the channel from the received signal of each channel
First sound detectors 1 for the number of channels for detecting sound every time
06 and the second for detecting sound from the microphone input signal
Voice detection unit 107, the first voice detection unit 106 and the
2 by the voice detection by the voice detection unit 107 of FIG.
And there is no sound in the microphone input signal,
Add uncorrelated signal between channels to received signal of channel
An uncorrelated signal adding unit 102, to which the uncorrelated signal is added
Speaker 103 using the received signal₁~ 103_mAnd microphone
Estimate echo paths for the number of channels between the lophones 104
Adaptive filter 101, adapted from microphone input signal
Subtraction unit for reducing a pseudo echo output from the filter 101
105, the sound quality of the received signal is inferior.
Without changing the echo path between the speaker and the microphone.
Estimate and cancel echo.

Further, in the multi-channel echo canceling method according to the first embodiment of the present invention, the step of reproducing the received signals of each channel for the number of channels for each channel includes the steps of: , A step S401 for detecting a sound for each channel, a step S402 for detecting a sound from a microphone input signal, and a step S40.
1 and step S403, when there is no voice in the received signal and the microphone input signal, a signal uncorrelated between channels is added to the received signal of each channel.
In steps S401 and S402, when there is no sound in the received signal and the microphone input signal, the speaker 10 is used by using the received signal to which the decorrelation signal is added.
3 ₁ 10 @ 2 to 10 @ 3 _m and the step S404 of estimating the echo path for the number of channels between the microphone 104, since a step S405 to reduce the pseudo echo outputted from the microphone input signal at step S404, the sound quality of audio reception signal The echo path between the speaker and the microphone can be estimated and the echo can be canceled without deterioration.

In the above-described embodiment, as a program for controlling the operation of the multi-channel echo canceller, control means (not shown, for example, C
Although the description has been given of the case where the program stored in the PU and the ROM is used, the present invention relates to a recording medium (for example, a CD-ROM) readable by the personal computer in addition to the program stored in the control unit.
) Stored in the above-described steps S401 to S40.
The same effect can be obtained by using a program for executing the operation including the operation No. 5. In this case, the multi-channel echo canceling device, the storage medium, the control means, and the like of the above-described embodiment form a voice communication system.

[Second Embodiment] FIG. 5 shows a second embodiment of the present invention.
2 is a block diagram of a main part of the embodiment. This is different from the first embodiment in that first band splitters 108 _{1 to} 108 _m that further divide a received signal of each channel into a plurality of bands, and a second band splitter that divides a microphone input signal into a plurality of bands. A band dividing unit 109, and a third band dividing unit 11 for dividing a received signal of each channel to which the uncorrelated signal is added into a plurality of bands.
0 _{1 to} 110 _m , a band synthesizing unit 111 for synthesizing a band-divided signal after echo cancellation, and an uncorrelated signal adding unit 102 for adding a band-limited uncorrelated signal to a band-divided received signal. The provided point is different. According to this configuration, it is also possible to obtain an effect of estimating an echo path for each subband and eliminating echoes without deteriorating the sound quality of the voice of the received signal. Note that this other configuration is different from that of the first embodiment (shown in FIG. 1) in each of the divided bands.
Therefore, the same components are denoted by the same reference numerals and description thereof is omitted.

In FIG. 5, the bold lines show signals of the number of channels of the received signal. Also, the first band division unit 108
_i divides a reception signal of channel number i (i = 1 to m) into a plurality of bands b using a technique such as a DFT filter bank to obtain a subband reception signal. First band division unit 10
8 _i are provided for the number of channels of the received signal.

The second band division unit 109 divides the microphone input signal into a plurality of bands b in the same manner as the first band division unit 109, and obtains a sub-band microphone input signal.

The third band dividing section 110 _i of the first band dividing unit 108 _i a similar manner, the received signal a plurality of uncorrelated signals added channel number i (i = 1~m) The received signal is divided into band b and a sub-band uncorrelated signal is added. The third band division units 110 _i are provided for the number of channels of the received signal.

The band synthesizing unit 111 synthesizes the band-divided signals after echo cancellation to calculate a band synthesized signal (full band signal).

The uncorrelated signal adding section 102 sets the band number j when the first voice detecting section 106 _ij and the second voice detecting voice detecting section 107 _j of the channel number i and the band number j both determine that there is no voice. A signal that is uncorrelated between the channels whose band is limited to the above band is added to the subband reception signal of each channel.

Next, the operation principle of this embodiment will be described with reference to FIG. Note that in FIG. 6, steps S602, S603, S607, and S608 are performed.
Indicates a process of performing processing for each subband.

First, the first band division section 108 _i divides the reception signal of the channel number i into a plurality of bands b to obtain a sub-band reception signal (step S601).

Next, the first voice detection unit _106ij detects whether or not there is voice in the received signal of the channel number i and the band number j (step S602).

Next, in the second band division section 109, the microphone input signal is divided into a plurality of bands b to obtain a sub-band microphone input signal (step S60).
3).

[0038] Then, in the second speech detection unit 107 _j, detects whether there is a voice in the microphone input signal of the band number j (step S604).

Next, in the decorrelation signal adding section 102, the first voice detecting section 106 _ij and the second voice detecting section 1
07 If the sound is not detected by _j, uncorrelated to the received signal of the band number j between channels (independent) adding a signal (step S605).

Next, in the third band division section 110 _i , the reception signal to which the uncorrelated signal of channel number i is added is divided into a plurality of bands b, and a subband reception signal to which the uncorrelated signal is added is obtained. (Step S606).

Next, the sub-band adaptive filter 101 _j
In the case where no voice is detected by the first voice detection unit 106 _{i -j} and the second voice detection unit 107 _j , the subband reception signal added with the uncorrelated signal and the subband microphone input signal Estimate the echo path and calculate the sub-band pseudo echo (step S60)
7).

[0042] Then, in the subtraction unit 105, from the subband microphone input signal, by subtracting the sub-band pseudo echo, which is the output of the sub-band adaptive filters 101 _j, erases the sub-band echo (step S6
08).

Next, the band combining section 111 combines the sub-band microphone input signals after the echo cancellation,
A band synthesized signal (full band signal) is calculated (step S609).

As described above, in the multi-channel echo canceling method according to the second embodiment of the present invention, the steps of reproducing the received signals of each channel for the number of channels, for each channel,
Step S601 of dividing the received signal of each channel by the number of channels into a plurality of frequency bands, and the number obtained by multiplying the number of channels by the number of band divisions from the received signal of each channel, S602, the microphone input signal is divided into bands, step S603, the sound of the band is detected from the divided microphone input signal by the number of divided bands, and the number of channels to which the uncorrelated signal is added. Dividing the received signal of each channel into a plurality of frequency bands, step S602, and step S602.
When there is no sound in the received signal and the microphone input signal by the sound detection in an arbitrary band according to 604, steps S605 and S602 of adding an uncorrelated signal between the channels to the received signal of each channel in the band. Estimating an echo path for the number of channels between the speakers 103 _{1 to} 103 _m and the microphone 104 using the uncorrelated signal when there is no sound in the received signal and the microphone input signal by the sound detection in the arbitrary band in S604. S607, a step S608 of subtracting the sub-band pseudo echo output in step S607 from the sub-band microphone input signal, and a step S609 of synthesizing the band-divided sub-band echo-eliminated sub-band signal.
Therefore, the echo path can be estimated for each sub-band and the echo can be eliminated without deteriorating the sound quality of the voice of the received signal.

Further, in the above-described embodiment, as a program for controlling the operation of the multi-channel echo canceller, control means (not shown, such as C
Although the description has been given of the case where the program stored in the PU and the ROM is used, the present invention relates to a recording medium (for example, a CD-ROM) readable by the personal computer in addition to the program stored in the control unit.
) Stored in the above-described steps S601 to S60.
The same effect can be obtained by using a program for executing the operation including the operation No. 9. In this case, the multi-channel echo canceling device, the storage medium, the control means, and the like of the above-described embodiment form a voice communication system.

In the first and second embodiments, the case where the number of microphones is one has been described. However, the present invention is not limited to the case where the number of microphones is one. By preparing any of the above for the number of microphones, the same effect can be obtained even when applied to a plurality of microphones.

[0047]

As described above, the present invention reproduces the received signals of each channel corresponding to the number of channels for each channel, and detects the sound for each channel from the received signals of each channel corresponding to the number of channels. Sound is detected from the microphone input signal, and by detecting these sounds, when there is no sound in the received signal and the microphone input signal, an uncorrelated signal is added to the received signal of each channel between the channels. Estimating the number of echo paths for the number of channels between the speaker and the microphone using the added received signal, and reducing the pseudo echo output when estimating the echo path from the microphone input signal, the sound quality of the sound of the received signal is reduced. Multi-channel echo canceller, method, which has an excellent effect of estimating the echo path without deteriorating and canceling the echo, It is capable of providing recording media and voice communications systems.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a multi-channel echo canceling apparatus according to a first embodiment of the present invention.

FIG. 2 is a flowchart showing an operation of a first voice detection unit of the multi-channel echo canceling apparatus according to the first embodiment of the present invention.

FIG. 3 is a flowchart showing an operation of a second voice detection unit of the multi-channel echo canceling apparatus according to the first embodiment of the present invention.

FIG. 4 is a flowchart showing the operation of the multi-channel echo canceling apparatus according to the first embodiment of the present invention;

FIG. 5 is a block diagram showing a configuration of a multi-channel echo canceling apparatus according to a second embodiment of the present invention.

FIG. 6 is a flowchart showing the operation of the multi-channel echo canceling apparatus according to the second embodiment of the present invention.

FIG. 7 is a block diagram showing a configuration of a conventional multi-channel echo canceller.

[Explanation of symbols]

Reference Signs List 101 adaptive filter 102 uncorrelated signal addition unit 103 _i speaker 104 microphone 105 subtraction unit 106 _i first audio detection unit 107 second audio detection unit 108 _i first band division unit 109 second band division unit 110 _i 3 band dividing section 111 band synthesizing section S401 first speech detection step S402 second speech detection step S403 uncorrelated signal addition step S404 echo path estimation step S405 echo cancellation step S601 first band division step S602 first speech detection Step S603 Second band division step S604 Second voice detection step S605 Uncorrelated signal addition step S606 Third band division step S607 Echo path estimation step S608 Echo cancellation step S609 Band synthesis step

Claims (5)

[Claims] 1. A loudspeaker corresponding to the number of channels for reproducing a reception signal of each channel for each channel, and a first number of loudspeakers corresponding to the number of channels for detecting sound for each channel from the reception signal of each channel.
And a second sound detection unit for detecting sound from the microphone input signal, and sound detection by the first sound detection unit and the second sound detection unit, thereby converting the reception signal and the microphone input signal into sound. When there is no signal, an uncorrelated signal adding unit that adds a signal that is uncorrelated between channels to a received signal of each channel, and using the received signal to which the uncorrelated signal is added, the number of channels between the speaker and the microphone is reduced. A multi-channel echo canceller, comprising: an adaptive filter for estimating an echo path of the input signal; and a subtraction unit for subtracting a pseudo echo output from the adaptive filter from a microphone input signal. 2. A loudspeaker corresponding to the number of channels for reproducing a received signal of each channel for each channel, a first band dividing unit corresponding to the number of channels for dividing the received signal of each channel into a plurality of frequency bands, A first audio detector for detecting an audio of an arbitrary band of an arbitrary channel from the received signal of each channel obtained by multiplying the number of channels by the number of band divisions, and a second band for dividing a microphone input signal into bands A dividing unit, and a second sound detecting unit corresponding to the number of band divisions for detecting sound in the band from the microphone input signal subjected to the band division;
A third band division unit corresponding to the number of channels for dividing the reception signal of each channel to which the uncorrelated signal is added into a plurality of frequency bands, and an arbitrary band by the first audio detection unit and the second audio detection unit By the sound detection in the above, when there is no sound in the received signal and the microphone input signal, an uncorrelated signal adding unit for the number of band divisions that adds uncorrelated signals between channels to the received signal of each channel of the band, A sub-band adaptive filter for estimating an echo path for the number of channels between the speaker and the microphone using the uncorrelated signal; and a sub-band for subtracting a sub-band pseudo echo output from the sub-band adaptive filter from a sub-band microphone input signal. A multi-channel E / E unit comprising a subtraction unit and a band synthesis unit for synthesizing the sub-band signals after band-division echo cancellation. Over erasing device. 3. A signal reproducing procedure for reproducing, for each channel, received signals of each channel for the number of channels, a first audio detecting procedure for detecting audio for each channel from the received signals of each channel for the number of channels, A second voice detection procedure for detecting voice from a microphone input signal, and voice detection by the first voice detection procedure and the second voice detection procedure, when there is no voice in the received signal and the microphone input signal, A decorrelation signal addition procedure for adding a decorrelation signal between channels to a reception signal of a channel, and an echo path estimation for estimating an echo path for the number of channels between the speaker and the microphone using the reception signal to which the decorrelation signal is added And a subtraction procedure for subtracting a pseudo echo output in the path estimation procedure from a microphone input signal. Echo cancellation method. 4. A signal reproducing procedure for reproducing, for each channel, reception signals of each channel corresponding to the number of channels in a computer, and a first sound detection for detecting sound for each channel from the reception signals of each channel, corresponding to the number of channels. Procedures, a second speech detection procedure for detecting speech from the microphone input signal,
When there is no voice in the received signal and the microphone input signal by the voice detection by the first voice detection procedure and the second voice detection procedure, a non-correlated signal is added to the received signal of each channel between the channels. A correlation signal adding procedure,
An echo path estimation procedure for estimating the number of echo paths for the number of channels between a speaker and a microphone using the reception signal to which the uncorrelated signal is added, and an echo cancellation procedure for subtracting a pseudo echo output from the microphone input signal in the echo path estimation procedure And a computer-readable recording medium on which a program for executing the above is recorded. 5. A multi-channel echo canceling device according to claim 1 or 2, and control means for controlling the multi-channel echo canceling device by a program recorded on a recording medium according to claim 4. A voice communication system characterized by the following.