JP2012039441A - Multi-channel echo erasure method, multi-channel echo erasure device, and program of the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a multi-channel echo erasure technology which can erase an echo swiftly without being affected by variance in an echo path and cross-correlation between a double-talk and a receiving signal channel.SOLUTION: A residual echo erasure part or a second residual echo erasure part estimates an input-output transfer characteristic by using a power spectrum of a pseudo echo signal and a cross spectrum between a second sound collection signal or a third sound collection signal and the pseudo echo signal, and predicts a residual echo component included in the second sound collection signal or the third sound collection signal from the pseudo echo signal and the input-output transfer characteristic.

Description

  In the present invention, M speakers (M is an integer of 2 or more) and N microphones (N is an integer of 1 or more) are arranged in a common sound field, and M-channel received signals are reproduced from the speakers and collected by the microphone. The present invention relates to a technique for erasing an acoustic echo (hereinafter simply referred to as “echo”) from a collected sound signal (hereinafter referred to as “first sound collection signal”), and more particularly to a technique for eliminating echo in a loudspeaker communication system such as a video conference system.

  The received signal is reproduced by the speaker, and the sound is picked up by the microphone to generate an echo. If it is transmitted as it is, problems such as troubles in communication and discomfort arise. Further, howling occurs when the volume of the speaker or microphone is high, making it impossible to make a call. In particular, such a problem becomes conspicuous in the voice call system.

  In order to solve this problem, there is an echo canceller that cancels echoes using an adaptive filter as a prior art. Non-Patent Document 1 is known as a conventional multi-channel echo canceller. A conventional multi-channel echo canceller 90 will be described with reference to FIG.

M speakers ₂ 1, ..., _{2 M} and N microphones ₃ 1, ..., _{3 N} are arranged in a common sound field, speaker ₂ 1, ..., received signal from the _{2 M} M channels _x 1 (k ),..., X _M (k), the multi-channel echo canceller 90 cancels the playback sound (echo) that wraps around the microphone via the m × n echo paths h _mn (k). However, m = 1,..., M, n = 1,..., N, M is 2 or more, and N is 1 or more. Multi-channel echo canceller 90, transmitter terminal 4 ₁ of total N-channel of the receiving terminal ₁ 1 to 1 _M, the transmitting end of all M channels of the receiving side, ..., _{4 N} and a microphone ₃ 1, ..., _{3 N} There are connected, the received signal _{x 1 (k), ...,} x M (k) and the sound collection signal _{y 1 (k), ...,} y N (k) is input, transmission signal _u 1 (k) , _{..., u} N (k) a transmitter terminal ₄ 1, ..., and outputs a _{4 N.} The multi-channel echo canceller 90 includes N echo cancelers 8 ₁ ,..., 8 _N , and each echo canceller includes an echo predictor 81, a subtractor 82, and an echo path estimator 83. Figure 1 describes the echo cancellation unit _{8 1, y} 1 (k) of the y _{(k), u} 1 (k) of the _{u (k), h 11 (} k), ..., h M1 (k) of Expressed as h ₁ (k),..., h _M (k). The same processing can be performed on the collected sound signals from other microphones, and it is only necessary to arrange the configurations in FIG. 1 in parallel.

The multi-channel echo canceller 90 obtains a difference between the collected sound signal y (k) and the pseudo echo signal y ′ (k), that is, an error signal u (k) by the subtracting unit 82, and receives this signal u (k) and the received signal. The echo path estimator 83 sequentially estimates the echo path (filter coefficient h ′ (k)) from the signals x ₁ (k),..., X _M (k), and the echo predictor 81 uses the estimation result to simulate. An echo signal y ′ (k) is generated. In a state where the echo path estimation is performed with high accuracy, the echo component included in the collected sound signal y (k) and the pseudo echo signal y ′ (k) are almost equal, and the error signal u (k) includes almost no echo. It will not be.

M.M.Sondhi, D.R.Morgan, and J.L.Hall, “Stereophonic Acoustic Echo Cancellation-An Overview of the Fundamental Problem”, IEEE Signal Processing Letters, AUGUST 1995, vol.2, no.8, pp.148-151

  However, the prior art cannot always sufficiently cancel the echo.

  Since the echo path estimation by the echo path estimation unit 83 is not completed instantaneously, the residual echo increases every time the echo path changes due to human movement or the like.

  In the double talk state, since the error signal includes the voice of the sender, the estimation of the echo path is disturbed and the residual echo becomes large.

  Further, as described in Non-Patent Document 1, when the received signal is multi-channel, since the correlation between channels is high, the estimated echo path and the true path are true even when the echo is canceled. The echo paths may not always match. In that case, when the far-end speaker changes and the cross-correlation between the channels of the received signal changes, the residual echo suddenly increases.

  Then, there is a problem that the call quality deteriorates when the residual echo becomes large.

  In order to solve the above-described problems, the multi-channel echo cancellation technique according to the present invention includes a pseudo echo signal power spectrum, a second sound collection signal, or a third sound collection signal in the residual echo cancellation unit or the second residual echo cancellation unit. The input / output transfer characteristic is estimated using the cross spectrum between the sound signal and the pseudo echo signal, and the residual echo component contained in the second sound pickup signal or the third sound pickup signal is determined from the pseudo echo signal and the input / output transfer characteristic. Predict.

  Since the present invention predicts the residual echo component without using an adaptive filter in the residual echo canceller or the second residual echo canceler, it is not affected by echo path fluctuations, double talk, and cross-correlation between received signal channels. In addition, it is possible to quickly cancel the echo.

The figure which shows the structural example of the conventional multichannel echo cancellation apparatus 90. FIG. FIG. 2 is a diagram illustrating a configuration example of multi-channel echo cancellation apparatuses 100 and 400. The figure which shows the processing flow of the multi-channel echo cancellation apparatuses 100 and 400. The figure which shows the structural example of the input-output correlation coefficient calculation part 603. FIG. The figure which shows the structural example of the multichannel echo cancellation apparatuses 200 and 500. FIG. The figure which shows the processing flow of the multichannel echo cancellation apparatuses 200 and 500. The figure which shows the structural example of the input-output correlation coefficient calculation parts 2603 and 903. FIG. 3 is a diagram illustrating a configuration example of a multi-channel echo canceller 300. The figure which shows the processing flow of the multichannel echo cancellation apparatus 300. The figure which shows the structural example of the input-output correlation coefficient calculation part 3603. Diagram illustrating a second configuration example of the residual echo cancellation unit 39 _1. The figure which shows the structural example of the transfer characteristic adjustment parts 4604 and 5604. FIG.

  Hereinafter, embodiments of the present invention will be described in detail.

<Multi-channel echo canceller 100>
A multi-channel echo canceling apparatus 100 according to the first embodiment will be described with reference to FIGS. 2 and 3. Multi-channel echo canceller 100, N pieces of echo cancellation portion _81, ..., includes a _{8 N,} N pieces of the residual echo cancellation portion _61, ..., a _{6 N.}

M speakers ₂ 1, ..., _{2 M} and N microphones ₃ 1, ..., _{3 N} are arranged in a common sound field, speaker ₂ 1, ..., received signal from the _{2 M} M channels _x 1 (k ),..., X _M (k) is reproduced, the multi-channel echo canceller 100 cancels the reproduced sound (echo) that wraps around the microphone via the m × n echo paths h _mn (k). However, m = 1,..., M, n = 1,..., N, M is 2 or more, and N is 1 or more. Multi-channel echo canceller 100, transmitter terminal ₄ 1 of total N-channel of the receiving terminal ₁ 1 to 1 _M, the transmitting end of all M channels of the receiving side, ..., _{4 N} and a microphone ₃ 1, ..., _{3 N} There are connected, the received signal _{x 1 (k), ...,} x M (k) and the first collected signal _y 1 (k), _..., and inputs the _y N (k), transmission signal _v 1 ( _{k), ..., v N (} k) the transmitter terminal ₄ 1, ..., and output to _{4 N.}

It explained echo canceling portion ₈₁ and the residual echo canceling portion ₆₁ of FIG. Note that the same processing can be performed on the collected sound signals from other microphones in the echo canceling unit 8 _n and the residual echo canceling unit 6 _n . Therefore, it is only necessary to arrange the configuration of FIG. 2 in parallel. In the following drawings, y ₁ (k) is y (k), v ₁ (k) is v (k), and h ₁₁ (k),..., H _M1 (k) is h ₁ (k). ,..., H _M (k).

<Echo elimination unit 8 ₁ >
Echo canceling portion ₈₁ is provided with an echo prediction unit 81, subtraction unit 82 and the echo path estimation unit 83.

First, the impulse response h _m of the echo path from the speaker 2 _m to the microphone 3 _{1 (k),} the length When L _E, between the received signal of the M-channel and the first collected signal y (k) Have the following relationship:

Impulse response and received signal of each channel
h _m = [h _m (0)… h _m (L _E -1)] ^T (2)
x _m = [x _m (k)… x _m (kL _E +1)] ^T (3)
As a vector, the relationship between the M channel received signal and the first collected sound signal is described as follows.

y (k) = h ₁ ^T x ₁ (k) +… + h _M ^T x _M (k) (4)
However, T represents transposition.

<Echo Prediction Unit 81>
The echo prediction unit 81 inputs the received signals x ₁ (k),..., X _M (k) to the pseudo echo path by the adaptive filter, generates and outputs a pseudo echo signal y ′ (k) (s81). The echo prediction unit 81 includes an adaptive filter, and the characteristic of the adaptive filter is controlled by an echo path estimation unit 83 (to be described later) so that the error signal of the subtraction unit 82 in the reception state is minimized.

For example, the filter coefficient of the adaptive filter for each channel
h ' _m = [h' _m (0)… h ' _m (L _E -1)] ^T (5)
age,
y '(k) = h' ₁ ^T x ₁ (k) +… + h ' _M ^T x _M (k) (6)
Is generated. Echo prediction unit 81 includes a pseudo echo signal y generated '(k) to the subtraction unit 82, and outputs it to the frequency domain transform section 608 of the residual echo canceling unit 6 _1. For example, the tap length of the adaptive filter may be set to about 100 to 200 ms.

<Subtraction unit 82>
The subtracting unit 82 receives the first sound pickup signal y (k) and the pseudo echo signal y ′ (k), subtracts the pseudo echo signal y ′ (k) from the first sound pickup signal y (k), and generates an error signal. U (k) (hereinafter referred to as “second sound pickup signal”) is obtained and output (s82).

u (k) = y (k) -y '(k) (7)
The obtained second collected sound signal u (k) is output to the echo path estimation unit 83 and the frequency domain conversion unit 602 in the residual echo cancellation unit 61.

<Echo path estimation unit 83>
The echo path estimation unit receives the second collected sound signal u (k) and the received signals x ₁ (k),..., X _M (k) as input, and uses them to use the filter coefficient h ′ (k) of the adaptive filter. Is updated and output (s83). A case where the Normalized Least Mean Square algorithm (NLMS algorithm) is used as a coefficient correction method for the adaptive filter will be described.

h ' _m (k + 1) = h' _m (k) + μu (k) x _m (k) (8)
Where μ is the step size,

Determined by. Note that μ ₀ is a parameter that is controlled based on the power of the input signal and is preset to a value of 0 to 1 in order to perform stable estimation. The echo path estimation unit 83 copies the updated filter coefficient h ′ (k + 1) and outputs it to the echo prediction unit 81.

<Residual echo canceller 6 ₁ >
Residual echo canceling unit _{6 1,} M number of frequency domain transform unit ₆₀₁ 1, ..., and 601 _M, the frequency domain converter 602, an input-output correlation coefficient calculation unit 603, an input-output transfer characteristic estimating section 604, It has a residual echo prediction unit 605, a subtraction unit 606, a time domain conversion unit 607, and a frequency domain conversion unit 608.

<Frequency domain transform units 601 ₁ ,..., 601 _M , 602, 608>
The frequency domain transform units 601 ₁ ,..., 601 _M each receive M channel received signals x ₁ (k),..., X _M (k) as inputs, and frequency domain signals X ₁ (f, j),. X _M (f, j) is converted and output (s601). However, f represents a frequency number and j represents a frame number.

  The frequency domain transforming unit 602 receives the second collected sound signal u (k), which is an output signal of the subtracting unit 82, converts it into a frequency domain signal U (f, j), and outputs it (s602).

The frequency domain transform unit 608 receives the pseudo echo signal y ′ (k), which is an output signal of the echo prediction unit 81, converts it into a frequency domain signal, and outputs it (s608). Note that the frequency domain pseudo echo signal is represented as X ₀ (f, j) for convenience.

  A case will be described in which each signal is set to 1 frame = 2 L samples, the L / D samples are blocked, and the L / D samples are shifted to create a frame. However, L is a natural number, D is a natural number that divides L, and time k = jL / D. The conversion to the frequency domain is performed by, for example, FFT (Fast Fourier transform) or DFT (discrete Fourier transform), and L may be a power of 2 in order to simplify and speed up the calculation. For example, L = 64 to 512, D = 2 to 8 and the like. The frame length may be set so as to correspond to 10 ms to 20 ms.

<Input / output correlation coefficient calculation unit 603>
Output phase number calculating section 603 relationship, the pseudo echo signal _X 0 in the frequency domain (f, j) and the received signal of the M-channel _{X 1 (f, j),} ..., X M (f, j) and the second The collected sound signal U (f, j) is input, and the cross spectrum between these signals and the power spectrum of each channel of the pseudo echo signal and the M channel received signal are obtained and output (s603). Each cross spectrum and power spectrum are values at time k = jL / D.

  For example, the input / output correlation function calculation unit 603 includes a power spectrum calculation unit 603a and a cross spectrum calculation unit 603b as shown in FIG.

The power spectrum calculation unit 603a obtains the power spectrum P ₀₀ (f, j) of the pseudo echo signal and the power spectrum P _mm (f, j) of the m-th channel received signal of the M channel received signal by the following equations. .

However, q = 0, 1,..., M, A ^* represents a complex conjugate of A, and E [A] represents a function that averages A.

The cross spectrum calculation unit 603b obtains the cross spectrum P _0m (f, j) between the pseudo echo signal and the m-th channel received signal, the m′-channel received signal, and the m-th channel received by the following equations. P _m′m (f, j) is obtained from the cross spectrum between the signals. Note that m ′ ≠ m, and m ′ = 1,.

However, q ′ ≠ q and q ′ = 0, 1,.

Further, the cross spectrum calculation unit 603b calculates the cross spectrum Q ₀ (f, j) between the pseudo echo signal and the second sound collection signal, the m-th channel reception signal, and the second sound collection signal according to the following equations. A cross spectrum Q _m (f, j) between the two is obtained.

As an example of the averaging process,

As described above, there is a method using a processing result of one frame before and a smoothing constant β that takes a value of 0 to 1. Further, a method of obtaining an average by multiplying the past several frames by a time constant is conceivable. The same applies to Q (f, j).

The input / output correlation coefficient calculation unit 603 includes a correlation coefficient P (f, j) between input signals including the power spectrum P _qq (f, j) and the cross spectrum P _q′q (f, j), In other words, the correlation coefficient Q (f, j) between the input and output signals is obtained.

<Input / output transfer characteristic estimation unit 604>
The input / output transfer characteristic estimator 604 includes the pseudo echo signal and the power spectrum P _qq (f, j) of each M-channel received signal, and between the pseudo echo signal, each M-channel received signal, and the second collected sound signal. Using the cross spectrum P _q′q (f, j) and Q _q (f, j), the input / output transfer characteristic G (f, j) is estimated for each frequency (s604) and output.

  For example, the input / output transfer characteristic estimation unit 604 calculates the input / output transfer characteristic G (f, j).

Estimated by

  For the matrix composed of the power spectrum and the cross spectrum, a small constant may be added to the diagonal component in order to stabilize the inverse matrix calculation.

<Residual echo prediction unit 605>
The residual echo prediction unit 605 estimates the input / output transfer characteristics estimated as the M-channel received signals X ₁ (f, j),..., X _M (f, j) and the pseudo echo signal X ₀ (f, j) in the frequency domain. Using G (f, j) as an input, a residual echo component U ^ (f, j) included in the second collected sound signal is predicted from these values (s605) and output. For example, the residual echo prediction unit 605 predicts the residual echo component by the following formula.

<Subtraction unit 606 and time domain conversion unit 607>
The subtraction unit 606 receives the second collected sound signal U (f, j) and the predicted residual echo component U ^ (f, j) as input, and predicts from the second collected sound signal U (f, j) in the frequency domain. The transmitted echo signal U (f, j) is subtracted (s606), and the transmission signal V (f, j) is obtained and output.

V (f, j) = U (f, j) -U ^ (f, j) (19)
The time domain conversion unit 607 receives the transmission signal V (f, j) obtained by the subtraction unit 606, converts it into a time domain signal v (k), and outputs it to the transmission terminal 41 as a transmission signal. (S607). Note that the conversion to the time domain may be performed as long as it corresponds to the conversion method used in the frequency domain conversion unit, for example.

<Effect>
With such a configuration, it is possible to maintain the prior art equivalent amount of echo cancellation and voice quality by echo cancellation unit 8 _n, further comprising a power spectrum of each channel of the received signals of the pseudo echo signal and the M-channel The input / output transfer characteristics are estimated from the pseudo echo signal, the cross spectrum between each channel of the M channel received signal and the second sound pickup signal, and the M channel received signal, the pseudo echo signal, and the input / output transfer characteristics are Therefore, since the echo component included in the second collected sound signal is predicted, the residual echo canceling unit does not need to update the filter coefficient of the adaptive filter. Therefore, it is possible to respond immediately to fluctuations in the echo path, predict an echo faster than conventional multi-channel echo cancellers, and suppress estimated fluctuations due to signals other than echoes. In addition, since the amplitude and phase of the transfer characteristics and echoes are estimated in the frequency domain and the echo is canceled by subtraction, the distortion of transmission loss can be reduced even during double talk, and the echo can be canceled regardless of the number of channels. it can.

That is, in the case where the initial stage or the situation where it takes time to estimate the echo path in the echo canceling unit 8 _n changes, the residual echo can be suppressed by performing the echo canceling processing by the high-speed residual echo canceling unit 6 _n. . Further, when the echo path estimation in the echo cancellation unit is stabilized it can be suppressed more residual echo by performance of the echo cancellation unit 8 _n. Therefore, compared with the case where the echo canceling unit and the residual echo canceling unit are used alone, it is possible to perform a call with the residual echo reduced over the entire processing time.

Further, in the residual echo canceller 6 _n , the processing delay amount is determined by L / D set by the frequency domain converters 601 _m , 602, and 608. If the frame length is increased in order to improve the prediction performance, the amount of delay increases. On the other hand, when shorter than tap length L _E of the adaptive filter frame length in order to shorten the processing delay (used in the echo prediction unit 81), can not be corresponding to the residual echo component reaching later than the frame length of the reverberation component . As a result, the residual echo cancellation performance is degraded. For example, when the frame length is 10 ms, the reverberation time of a normal conference room is 300 ms or more, and therefore it is not possible to cope with the residual echo component included in the portion after 10 ms of the echo path impulse response (that is, 10 ms to 300 ms or more). In addition, the residual echo cancellation performance is significantly degraded.

  Therefore, attention is paid to the fact that the pseudo echo signal y ′ (k) generated by the adaptive filter includes a reverberation component exceeding the frame length. In the residual echo prediction unit 605, the residual echo cancellation performance can be improved without increasing the delay amount by estimating the residual echo using the pseudo echo signal y '(k). Thereby, the residual echo cancellation performance can be ensured even in a room with long reverberation.

<Program>
The above-described multi-channel echo canceller can also be operated by a computer. In this case, each process of the program for causing the computer to function as the target device (the device having the functional configuration shown in the drawings in the embodiment) or the processing procedure (shown in the embodiment) is executed on the computer. The program may be downloaded into a computer from a recording medium such as a CD-ROM, a magnetic disk, or a semiconductor storage device or via a communication line, and the program may be executed.

<Modification>
Received signals _X 1 of M channels in the frequency space used by each residual echo cancellation portion (f, j), _..., the value of X M (f, j), the residual echo canceling portion _61, ..., in _{6 N} Therefore, N residual echo cancellers 6 ₁ ,..., 6 _N may share the output of the frequency domain transform unit of the received signal.

  Even if echo cancellation processing is not performed for all transmission signals, echoes may be canceled only for some transmission signals.

The input to the multi-channel echo canceller 100, receiving terminal 1 _1, ..., to the signal obtained from the 1 _M, may be a received signal subjected to correlation change treatment.

  As a method for updating the filter coefficient in the echo path estimation unit 83, a conventional technique other than the learning identification method (for example, a projection algorithm, an exponential weighting algorithm, an exponential weighted projection algorithm, or the like) may be used.

  The calculation amount of the correlation coefficient and the calculation of the transfer characteristic occupy most of the calculation amount of the residual echo canceller in the first embodiment. The calculation of the correlation coefficient is proportional to the square of the number of channels, and the calculation of the transfer characteristic is proportional to the third power of the number of channels because it includes an inverse matrix operation. For this reason, in Example 1, when the number of channels increases, the amount of calculation increases rapidly. Therefore, in the second embodiment, the configuration is simplified in order to suppress an increase in the amount of calculation when the number of channels increases.

<Multi-channel echo canceller 200>
A multi-channel echo canceling apparatus 200 according to the second embodiment will be described with reference to FIGS. 5 and 6. Only parts different from the first embodiment will be described. Multi-channel echo canceller 200, N pieces of echo cancellation portion _81, ..., includes a _{8 N,} N pieces of the residual echo cancellation unit ₂₆ 1, ..., a 26 _N.
With reference to FIG. 5, Example 1 is different from the residual echo canceling unit 26 ₁ will be described.

<Residual echo canceling unit 26 ₁ >
Residual echo canceling portion 26 ₁ includes a frequency domain transform unit 602, an input-output correlation coefficient calculation unit 2603, the input-output transfer characteristic estimating unit 2604, a residual echo prediction unit 2605, a subtracting unit 606, the time domain converter 607 and a frequency domain converter 608. Residual echo canceling portion 26 _1, M number of frequency domain transform unit ₆₀₁ 1, ..., it may not comprise a 601 _M.

<Input / output correlation coefficient calculation unit 2603>
The input / output correlation coefficient calculation unit 2603 receives the frequency domain pseudo echo signal X ₀ (f, j) and the second sound pickup signal U (f, j), and inputs the cross spectrum between these signals and the pseudo echo. The power spectrum of the signal is obtained and output (s2603). Each cross spectrum and power spectrum are values at time k = jL / D.

  For example, the input / output correlation function calculation unit 2603 includes a power spectrum calculation unit 2603a and a cross spectrum calculation unit 2603b as shown in FIG.

The power spectrum calculation unit 2603a obtains the power spectrum P ₀₀ (f, j) of the pseudo echo signal by the following equation.

The cross spectrum calculation unit 2603b obtains a cross spectrum Q ₀ (f, j) between the pseudo echo signal and the second sound collection signal by the following equation.

<Input / output transfer characteristic estimation unit 2604>
The input / output transfer characteristic estimation unit 2604 uses the power spectrum P ₀₀ (f, j) of the pseudo echo signal and the cross spectrum Q ₀ (f, j) between the pseudo echo signal and the second collected sound signal, The input / output transfer characteristic G ₀ (f, j) is estimated for each frequency (s2604) and output.

For example, the input / output transfer characteristic estimation unit 2604 calculates the input / output transfer characteristic G ₀ (f, j).

Estimated by

<Residual echo prediction unit 2605>
The residual echo prediction unit 2605 receives the estimated input / output transfer characteristic G ₀ (f, j) as the pseudo echo signal X ₀ (f, j), and based on these values, the residual echo component included in the second collected sound signal U ^ (f, j) is predicted (s2605) and output. For example, the residual echo prediction unit 2605 predicts the residual echo component by the following equation.

<Effect>
With this configuration, the residual echo canceling unit estimates the input / output transfer characteristics using the cross spectrum between the power spectrum of the pseudo echo signal and the second collected sound signal and the pseudo echo signal, Since the residual echo component contained in the second collected signal is predicted from the signal and input / output transfer characteristics, the echo is quickly canceled without being affected by echo path fluctuations, double-talk, and cross-correlation between channels of the received signal. There is an effect that can be done.

  Furthermore, even if the number of channels is increased, there is an effect that the calculation amount of the residual echo canceling unit is not increased. That is, since the input to the residual echo canceller is one channel (for pseudo echo signal), the amount of computation of the residual echo canceller does not increase even if the number of channels increases. Therefore, the configuration is particularly effective when the number of channels is large.

<Multi-channel echo canceller 300>
A multi-channel echo cancellation apparatus 300 according to the third embodiment will be described with reference to FIGS. 8 and 9. Only parts different from the first embodiment will be described. The multi-channel echo canceller 300 includes N echo cancelers 8 ₁ ,..., 8 _N , N first residual echo cancelers 36 ₁ ,..., 36 _N , and N second residual echo cancelers. 39 ₁ ,..., 39 _N are provided.

With reference to FIGS. 8 and 9, Example 1 is different from the first residual echo canceling portion 36 ₁ and the second residual echo canceling unit 39 ₁ will be described.

<First Residual Echo Elimination Unit 36 ₁ >
Residual echo canceling portion 36 _1, M number of frequency domain transform unit ₆₀₁ 1, ..., and 601 _M, the frequency domain converter 602, an input-output correlation coefficient calculation unit 3603, the input-output transfer characteristic estimating unit 3604, A residual echo prediction unit 3605 and a subtraction unit 3606 are included.

<Input / output correlation coefficient calculation unit 3603>
The input / output correlation coefficient calculation unit 3603 receives the M-channel received signals X ₁ (f, j),..., X _M (f, j) and the second collected sound signal U (f, j) in the frequency domain. Then, the cross spectrum between these signals and the power spectrum of each channel of the M channel received signal are obtained and output (s3603). Each cross spectrum and power spectrum are values at time k = jL / D.

  For example, the input / output correlation function calculation unit 3603 includes a power spectrum calculation unit 3603a and a cross spectrum calculation unit 3603b as illustrated in FIG.

The power spectrum calculation unit 3603a obtains the power spectrum P _mm (f, j) of the m-th channel received signal of the M channel received signal by the following equation.

The cross spectrum calculation unit 3603b obtains P _m′m (f, j) as a cross spectrum between the _m′th channel received signal and the mth channel received signal by the following equation.

Further, the cross spectrum calculation unit 3603b obtains a cross spectrum Q _m (f, j) between the m-th channel received signal and the second sound pickup signal by the following equation.

Note that the input / output correlation coefficient calculation unit 3603 has a correlation coefficient P (f, j) between input signals including a power spectrum P _mm (f, j) and a cross spectrum P _m′m (f, j) shown below. ) And the correlation coefficient Q (f, j) between the input and output signals can be paraphrased.

<Input / output transfer characteristic estimation unit 3604>
The input / output transfer characteristic estimator 3604 generates a power spectrum P _mm (f, j) of each M channel received signal and a cross spectrum P _{m ′} between the m′th channel received signal and the mth channel received signal. _{Using m} (f, j), the cross spectrum between each received signal of the M-th channel and the second collected sound signal, and Q _m (f, j), input / output transfer characteristics G (f, j j) is estimated (s3604) and output.

  For example, the input / output transfer characteristic estimation unit 3604 calculates the input / output transfer characteristic G (f, j).

Estimated by

<Residual echo prediction unit 3605>
The residual echo prediction unit 3605 receives the estimated input / output transfer characteristics G (f, j) as M-channel received signals X ₁ (f, j),..., X _M (f, j) in the frequency domain. The residual echo component U ^ (f, j) included in the second collected sound signal is predicted from the value of (2) (s3605) and output. For example, the residual echo prediction unit 3605 predicts the residual echo component by the following formula.

<Subtraction unit 3606>
The subtractor 3606 receives the second collected sound signal U (f, j) and the predicted residual echo component U ^ (f, j) as input, and predicts from the second collected sound signal U (f, j) in the frequency domain. ^ the residual echo component U (f, j) subtracted (S606), the third picked-up signal U '(f, j) and determined, and outputs the second to the residual echo canceling portion 39 _1.

    U '(f, j) = U (f, j) -U ^ (f, j) (38)

<Second Residual Echo Eliminating Unit 39 ₁ >
As shown in FIG. 11, the second residual echo canceling portion 39 ₁ includes a frequency domain converter 908, a second input-output correlation coefficient calculation unit 903, a second input-output transfer characteristic estimating section 904 second residual echo A prediction unit 905 and a subtraction unit 906 are provided. Second residual echo canceling unit 39 _1, the third picked-up signal U '(f, j) and the echo replica signal y' as input (k), further residual third collected signal U '(f, j) It generates a transmission signal excluding the echo component, and outputs the transmitting end 4 _1. Hereinafter, processing examples of each unit will be shown.

<Frequency domain transform unit 908>
The frequency domain transform unit 908 receives the pseudo echo signal y ′ (k) that is the output signal of the echo prediction unit 81, converts the pseudo echo signal y ′ (k) into a frequency domain signal X ₀ (f, j), and outputs the signal (s908).

<Second input / output correlation coefficient calculation unit 903>
The second input / output correlation coefficient calculation unit 903 receives the frequency domain pseudo echo signal X ₀ (f, j) and the third collected sound signal U ′ (f, j) as input, and the cross spectrum between these signals. Then, the power spectrum of the pseudo echo signal is obtained and output (s903). Each cross spectrum and power spectrum are values at time k = jL / D.

  For example, the second input / output correlation function calculator 903 includes a power spectrum calculator 903a and a cross spectrum calculator 903b as shown in FIG.

The power spectrum calculation unit 903a obtains the power spectrum P ₀₀ (f, j) of the pseudo echo signal by the following equation.

The cross spectrum calculation unit 903b obtains a cross spectrum Q ₀ (f, j) between the pseudo echo signal and the third sound collection signal by the following equation.

<Second input / output transfer characteristic estimation unit 904>
The second input / output transfer characteristic estimation unit 904 uses the power spectrum P ₀₀ (f, j) of the pseudo echo signal and the cross spectrum Q ₀ (f, j) between the pseudo echo signal and the third sound collection signal. The second input / output transfer characteristic G ₀ (f, j) is estimated for each frequency (s904) and output.

For example, the second input / output transfer characteristic estimation unit 904 calculates the second input / output transfer characteristic G ₀ (f, j).

Estimated by

<Second Residual Echo Prediction Unit 905>
The second residual echo predicting unit 905 receives the second input / output transfer characteristic G ₀ (f, j) estimated as the pseudo echo signal X ₀ (f, j), and is included in the third collected sound signal from these values. The second residual echo component U ^ '(f, j) to be generated is predicted (s905) and output. For example, the second residual echo prediction unit 905 predicts the second residual echo component by the following formula.

<Subtraction unit 906>
The subtraction unit 906 receives the third collected sound signal U ′ (f, j) and the predicted second residual echo component U ^ ′ (f, j) as input, and the third collected sound signal U ′ (f, j) in the frequency domain. The predicted residual echo component U ^ '(f, j) is subtracted from j) (s906), and the transmission signal V (f, j) is obtained and output.

    V (f, j) = U '(f, j) -U ^' (f, j) (45)

<Effect>
By adopting such a configuration, the same effect as in the first embodiment can be obtained. Furthermore, as will be described below, the amount of calculation can be reduced. In order to obtain the transfer characteristic related to the residual echo, the first embodiment uses Expression (17), and the third embodiment uses Expression (36) and Expression (43). In the expression (17) of the first embodiment, it is necessary to obtain an inverse matrix of (M + 1) × (M + 1), but in the corresponding part of the third embodiment, an M × M inverse matrix and scalar division are replaced. Since the calculation of the inverse matrix is proportional to the cube of the matrix size, the calculation amount for obtaining the transfer characteristic regarding the residual echo in the third embodiment is reduced to (M / (M + 1)) ³ times that in the first embodiment.

<Multi-channel echo canceller 400>
A multi-channel echo canceling apparatus 400 according to the fourth embodiment will be described with reference to FIGS. 2 and 3. Only parts different from the first embodiment will be described. Multi-channel echo canceller 400, N pieces of echo cancellation portion _81, ..., includes a _{8 N,} N pieces of the residual echo cancellation unit ₄₆ 1, ..., a 46 _N. A residual echo canceling unit different from that of the first embodiment will be described.

<Residual echo canceller 46 ₁ >
Residual echo canceling portion 46 _1, M number of frequency domain transform unit ₆₀₁ 1, ..., and 601 _M, the frequency domain converter 602, an input-output correlation coefficient calculation unit 603, an input-output transfer characteristic estimating section 604, The apparatus includes a residual echo prediction unit 605, a subtraction unit 606, a time domain conversion unit 607, and a frequency domain conversion unit 608, and further includes a transfer characteristic adjustment unit 4604. A transfer characteristic adjustment unit 4604 different from that of the first embodiment will be described.

<Transfer characteristic adjustment unit 4604>
As shown in FIG. 12, the transfer characteristic adjustment unit 4604 includes a magnitude calculation unit 4604a, a determination unit 4604b, and an adjustment unit 4604c, and the magnitude of the estimated input / output transfer characteristic G (f, j) is larger than the reference value. Is larger, the process of adjusting so that the magnitude of the input / output transfer characteristic G (f, j) matches the reference value, so-called clip processing is performed.

  The transfer characteristic adjustment unit 4604 receives the input / output transfer characteristic G (f, j), which is the output of the input / output transfer characteristic estimation unit 604, stores the input / output transfer characteristic G '(f, j). ) Or the input / output transfer characteristic G (f, j) as an input value is output as it is. Hereinafter, processing of each unit will be described.

  For example, the magnitude calculating unit 4604a calculates a norm | G (f, j) | as the magnitude of the estimated input / output transfer characteristic G (f, j). For example,

Asking.

  The determination unit 4604b determines whether or not the norm | G (f, j) | is greater than the reference value C specified in advance. When the norm | G (f, j) | is smaller than the reference value C, the input / output transfer characteristic G (f, j) as an input value is output as it is. When the norm is larger than the reference value, an adjustment instruction is transmitted to the adjustment unit 4604c. The reference value (value to be clipped) C is set as, for example, [an assumed acoustic coupling amount of one acoustic path] × [sqrt (number of acoustic paths)]. Further, for example, a value of about (1/3) to (1/2) of an assumed acoustic coupling amount of one acoustic path may be set as the reference value C.

  When receiving the adjustment instruction, the adjustment unit 4604c multiplies the input / output transfer characteristic G (f, j) by the reference value C and further divides by the norm | G (f, j) | f, j) is adjusted (s4604). For example, it is represented by the following formula.

The residual echo prediction unit 605 includes the estimated input / output transfer characteristic G (f, j) or the adjusted input / output transfer characteristic G ′ (f, j), and the M-channel received signal X ₁ (f, j) in the frequency domain. ,..., X _M (f, j) and the pseudo echo signal X ₀ (f, j) are input, and the residual echo component U ^ (f, j) included in the second sound pickup signal is predicted from these values. (S605) and output.

<Effect>
By adopting such a configuration, in addition to the same effects as in the first embodiment, the following effects can be obtained.

  By correcting the input / output transfer characteristic by performing the threshold comparison in this way, it is possible to prevent the input / output transfer characteristic from being overestimated due to the correlation between transmission and reception. Usually, there is no correlation between the voice spoken by the sender and the received signal. However, if the frame of the processing unit is shortened, a false correlation may occur between the voice spoken by the sender and the received signal. In this case, the input / output transfer characteristic is overestimated, and the predicted value of the residual echo component becomes larger than necessary, which may deteriorate the voice quality. In this embodiment, the deterioration can be prevented.

<Modification>
In the multi-channel echo canceller 300 according to the third embodiment, a transfer characteristic adjusting unit 4604 may be provided between the residual echo predicting unit 3605 after the input / output transfer characteristic estimating unit 3604. Even in this case, it is possible to prevent the input / output transfer characteristic from being overestimated.

  Further, the magnitude of the input / output transfer characteristic may be obtained by a conventional technique other than Expression (51).

<Multi-channel echo canceller 500>
A multi-channel echo canceling apparatus 500 according to the fifth embodiment will be described with reference to FIGS. Only parts different from the second embodiment will be described. Multi-channel echo canceller 500, N pieces of echo cancellation portion _81, ..., includes a _{8 N,} N pieces of the residual echo cancellation unit ₅₆ 1, ..., a 56 _N. A residual echo canceling unit different from that of the second embodiment will be described.

<Residual echo canceling unit 56 ₁ >
Residual echo canceling portion 56 ₁ includes a frequency domain transform unit 602, an input-output correlation coefficient calculation unit 603, an input-output transfer characteristic estimating unit 604, a residual echo prediction unit 605, a subtraction unit 606, the time domain converter 607 and a frequency domain conversion unit 608, and further includes a transfer characteristic adjustment unit 5604. A transfer characteristic adjustment unit 5604 different from that of the second embodiment will be described.

<Transfer characteristic adjustment unit 5604>
As shown in FIG. 12, the transfer characteristic adjustment unit 5604 includes a magnitude calculation unit 5604a, a determination unit 5604b, and an adjustment unit 5604c, and the magnitude of the estimated input / output transfer characteristic G ₀ (f, j) is a reference value. Is larger than the input / output transfer characteristic G ₀ (f, j), a process of adjusting so that the magnitude of the input / output transfer characteristic G ₀ (f, j) matches the reference value, so-called clip processing is performed.

The transfer characteristic adjustment unit 5604 receives the input / output transfer characteristic G ₀ (f, j), which is the output of the input / output transfer characteristic estimation unit 604, stores it, and adjusts the input / output transfer characteristic G ₀ ′ (f , J) or the input / output transfer characteristic G ₀ (f, j) as an input value is output as it is. Hereinafter, processing of each unit will be described.

The magnitude calculator 5604a calculates, for example, a norm | G ₀ (f, j) | as the estimated magnitude of the input / output transfer characteristic G ₀ (f, j). For example,

Asking.

The determination unit 5604b determines whether or not the norm | G ₀ (f, j) | is larger than the reference value C specified in advance. When the norm | G ₀ (f, j) | is smaller than the reference value C, the input / output transfer characteristic G ₀ (f, j) as an input value is output as it is. When the norm is larger than the reference value, an adjustment instruction is transmitted to the adjustment unit 5604c. The reference value (value to be clipped) C is set as, for example, [an assumed acoustic coupling amount of one acoustic path] × [sqrt (number of acoustic paths)]. Further, for example, a value of about (1/3) to (1/2) of an assumed acoustic coupling amount of one acoustic path may be set as the reference value C.

When the adjustment unit 5604c receives the adjustment instruction, the adjustment unit 5604c multiplies the input / output transfer characteristic G ₀ (f, j) by the reference value C, and further divides by the norm | G ₀ (f, j) | G ₀ (f, j) is adjusted (s5604). For example, it is represented by the following formula.

The residual echo prediction unit 2605 receives the estimated input / output transfer characteristic G ₀ (f, j) or the adjusted input / output transfer characteristic G ₀ ′ (f, j) and the pseudo echo signal X ₀ (f, j). The residual echo component U ^ (f, j) included in the second sound pickup signal is predicted from these values (s2605) and output.

<Effect>
By adopting such a configuration, in addition to the same effect as in the second embodiment, it is possible to prevent the input / output transfer characteristic from being overestimated due to the correlation between the transmission and the reception. Then, deterioration of voice quality can be prevented.

<Modification>
In the multi-channel echo canceller 300 of the third embodiment, a transfer characteristic adjusting unit 5604 may be provided between the second residual echo predicting unit 905 and subsequent to the second input / output transfer characteristic estimating unit 904. Even in this case, it is possible to prevent the input / output transfer characteristic from being overestimated.

  Further, the magnitude of the input / output transfer characteristic may be obtained by a conventional technique other than Expression (61).

100, 200, 300, 400, 500 Multi-channel echo cancellers 6 ₁ , 26 ₁ , 46 ₁ , 56 ₁ Residual echo canceller 36 ₁ First residual echo canceller 39 ₁ Second residual echo canceller 903 Second input / output Correlation coefficient calculation section 904 Second input / output transfer characteristic estimation section 905 Second residual echo prediction section 601 ₁ ,..., 601 _M , 602, 608, 908 Frequency domain conversion sections 603, 2603, 3603 Input / output correlation coefficient calculation section 604, 2604, 3604 Input / output transfer characteristic estimation unit 4604, 5604 Transfer characteristic adjustment unit 605, 2605, 3605 Residual echo prediction unit 606, 906 Subtraction unit 607 Time domain conversion unit 8 ₁ ,..., 8 _N echo cancellation unit 81 Echo prediction Unit 82 subtraction unit 83 echo path estimation unit

Claims (10)

M speakers (M is an integer equal to or greater than 2) and N microphones (N is an integer equal to or greater than 1) are arranged in a common sound field. A multi-channel echo canceling method for canceling echo that circulates into a microphone,
An echo prediction step of generating the pseudo echo signal by inputting the received signal to the pseudo echo path by the adaptive filter;
A subtraction step of subtracting the pseudo echo signal from a first sound pickup signal picked up by the microphone to obtain a second sound pickup signal;
An echo path estimating step of updating a filter coefficient of the adaptive filter using the second collected sound signal and the received signal;
A frequency domain conversion step of converting the pseudo echo signal and the second sound collection signal into frequency domain signals, respectively.
An input / output correlation coefficient calculating step for obtaining a cross spectrum between the power spectrum of the pseudo echo signal and the second sound collection signal and the pseudo echo signal;
An input / output transfer characteristic estimation step for estimating an input / output transfer characteristic for each frequency using a cross spectrum between a power spectrum of the pseudo echo signal, a second sound pickup signal, and the pseudo echo signal;
A residual echo prediction step for predicting a residual echo component included in the second sound pickup signal from the input / output transfer characteristic estimated as the pseudo echo signal in the frequency domain;
A subtracting step of subtracting a predicted residual echo component from the second collected sound signal in the frequency domain to obtain a transmission signal;
A time domain conversion step of converting the transmission signal obtained in the subtraction step into a time domain signal;
A multi-channel echo cancellation method comprising: The multi-channel echo cancellation method according to claim 1, comprising:
In the frequency domain conversion step, the pseudo echo signal, the second sound pickup signal, and the received signal are each converted into a frequency domain signal,
In the input / output correlation coefficient calculation step, a cross spectrum between the pseudo echo signal, each M-channel received signal and the second sound pickup signal, and each channel of the pseudo echo signal and the M channel received signal Power spectrum and
In the input / output transfer characteristic estimation step, a power spectrum of the pseudo echo signal and each M-channel received signal and a cross spectrum between the pseudo echo signal, each M-channel received signal and the second sound pickup signal are obtained. To estimate the input / output transfer characteristics for each frequency,
In the residual echo prediction step, a residual echo component included in the second collected sound signal is predicted from the input / output transfer characteristics estimated from the M channel reception signal and the pseudo echo signal in the frequency domain.
And a multi-channel echo cancellation method. The multi-channel echo cancellation method according to claim 1 or 2,
m ′ ≠ m, m = 1,..., M, m ′ = 1,..., M, q ′ ≠ q, q = 0, 1,. , M, the pseudo echo signal in the frequency domain is X ₀ (f, j), the m-th channel received signal is X _m (f, j), and the second sound pickup signal is U (f, j). The power spectrum of the pseudo echo signal is P ₀₀ (f, j), the power spectrum of the m-th channel received signal is P _mm (f, j), and the pseudo echo signal and the m-th channel received signal are the cross spectrum between the P _{0m (f,} j) and the cross spectrum between the received signal of the m 'channel and the reception signal of the m channels _P m'm _(f, j) and the pseudo echo signal When the cross spectrum between the second voice collecting signal and Q _{0 (f,} j), and the reception signal of the m-channel The cross spectrum between the two collected signals Q _{m (f,} j) and the A ^* is the complex conjugate of A, the E [A] as a function that takes an average of A,
In the input / output correlation coefficient calculating step, the power spectrum P ₀₀ of the pseudo echo signal and the power spectrum P _mm of the m-th channel received signal are calculated.

As calculated, the pseudo echo signal and the cross spectrum P _{0m (f,} j) between the received signal of the m channels and the cross spectrum P _m between the received signal of the m 'channel and the reception signal of the m-channel _'m (f, j)

And a cross spectrum Q ₀ (f, j) between the pseudo echo signal and the second sound pickup signal and a cross spectrum Q _m (f) between the m-th channel received signal and the second sound pickup signal. , J)

As sought
In the input / output transfer characteristic estimation step, the input / output transfer characteristic G (f, j) is

Estimated as
In the residual echo prediction step, the residual echo component is

Predict as
And a multi-channel echo cancellation method. M speakers (M is an integer equal to or greater than 2) and N microphones (N is an integer equal to or greater than 1) are arranged in a common sound field. A multi-channel echo canceling method for canceling echo that circulates into a microphone,
An echo prediction step of generating the pseudo echo signal by inputting the received signal to the pseudo echo path by the adaptive filter;
A subtraction step of subtracting the pseudo echo signal from a first sound pickup signal picked up by the microphone to obtain a second sound pickup signal;
An echo path estimating step of updating a filter coefficient of the adaptive filter using the second collected sound signal and the received signal;
A frequency domain conversion step of converting the pseudo echo signal, the second sound collection signal, and the reception signal into a frequency domain signal, respectively;
A first input / output correlation coefficient calculating step for obtaining a cross spectrum between each received signal of the M channel and the second collected sound signal and a power spectrum of each channel of the received signal of the M channel;
First, an input / output transfer characteristic is estimated for each frequency using a cross spectrum between each M-channel received signal and the second sound pickup signal and a power spectrum of each channel of the M-channel received signal. 1 I / O transfer characteristic estimation step;
A first residual echo prediction step of predicting a first residual echo component included in the second collected sound signal from the input / output transfer characteristics estimated from the M channel received signal in the frequency domain;
A subtracting step of subtracting the predicted first residual echo component from the second sound pickup signal in the frequency domain to obtain a third sound pickup signal;
A second input / output correlation coefficient calculating step for obtaining a power spectrum of the pseudo echo signal and a cross spectrum between the third sound pickup signal and the pseudo echo signal;
A second input / output transfer characteristic estimating step for estimating an input / output transfer characteristic for each frequency using a cross spectrum between a power spectrum of the pseudo echo signal, a third sound pickup signal, and the pseudo echo signal;
A second residual echo prediction step for predicting a second residual echo component included in the third sound pickup signal from the input / output transfer characteristic estimated as the pseudo echo signal in the frequency domain;
A subtraction step of subtracting the predicted second residual echo component from the third sound pickup signal in the frequency domain to obtain a transmission signal;
A time domain conversion step of converting the transmission signal obtained in the subtraction step into a time domain signal;
A multi-channel echo cancellation method comprising: A multi-channel echo cancellation method according to any one of claims 1 to 4,
A magnitude calculating step for calculating the magnitude of the estimated input / output transfer characteristic;
A determination step of determining whether or not the size of the input / output transfer characteristic is larger than a reference value designated in advance;
When the magnitude of the input / output transfer characteristic is larger than the reference value, the input / output transfer characteristic is multiplied by the reference value, and further divided by the magnitude of the input / output transfer characteristic. An adjustment step of adjusting the output transfer characteristic;
And a multi-channel echo cancellation method. M speakers (M is an integer equal to or greater than 2) and N microphones (N is an integer equal to or greater than 1) are arranged in a common sound field. A multi-channel echo canceller that cancels echo that circulates into the microphone,
An echo prediction unit that generates the pseudo echo signal by inputting the received signal into the pseudo echo path by the adaptive filter;
A subtractor for subtracting the pseudo echo signal from a first sound pickup signal picked up by the microphone to obtain a second sound pickup signal;
An echo path estimator that updates a filter coefficient of the adaptive filter using the second collected sound signal and the received signal;
A frequency domain conversion unit that converts the pseudo echo signal and the second sound collection signal into frequency domain signals, respectively;
An input / output correlation coefficient calculation unit for obtaining a power spectrum of the pseudo echo signal and a cross spectrum between the second sound pickup signal and the pseudo echo signal;
An input / output transfer characteristic estimator for estimating an input / output transfer characteristic for each frequency using a cross spectrum between the power spectrum of the pseudo echo signal, the second sound pickup signal, and the pseudo echo signal;
A residual echo prediction unit that predicts a residual echo component included in the second sound pickup signal from the input / output transfer characteristic estimated as the pseudo echo signal in the frequency domain;
A subtracting unit for subtracting the predicted residual echo component from the second collected sound signal in the frequency domain to obtain a transmission signal;
A time domain conversion unit for converting the transmission signal obtained by the subtraction unit into a time domain signal;
A multi-channel echo canceller. The multi-channel echo canceller according to claim 6,
In the frequency domain conversion unit, the pseudo echo signal, the second sound collection signal, and the reception signal are each converted into a frequency domain signal,
In the input / output correlation coefficient calculation unit, a cross spectrum between the pseudo echo signal, each M-channel received signal and the second sound pickup signal, and each channel of the pseudo echo signal and the M channel received signal Power spectrum and
In the input / output transfer characteristic estimation unit, a power spectrum of the pseudo echo signal and each of the M channel reception signals and a cross spectrum between the pseudo echo signal, each of the M channel reception signals and the second sound pickup signal are obtained. To estimate the input / output transfer characteristics for each frequency,
The residual echo prediction unit predicts a residual echo component included in the second collected sound signal from the input / output transfer characteristics estimated from the M-channel received signal and the pseudo echo signal in the frequency domain.
And a multi-channel echo canceller. The multi-channel echo canceller according to claim 6 or 7,
m ′ ≠ m, m = 1,..., M, m ′ = 1,..., M, q ′ ≠ q, q = 0, 1,. , M, the pseudo echo signal in the frequency domain is X ₀ (f, j), the m-th channel received signal is X _m (f, j), and the second sound pickup signal is U (f, j). The power spectrum of the pseudo echo signal is P ₀₀ (f, j), the power spectrum of the m-th channel received signal is P _mm (f, j), and the pseudo echo signal and the m-th channel received signal are the cross spectrum between the P _{0m (f,} j) and the cross spectrum between the received signal of the m 'channel and the reception signal of the m channels _P m'm _(f, j) and the pseudo echo signal When the cross spectrum between the second voice collecting signal and Q _{0 (f,} j), and the reception signal of the m-channel The cross spectrum between the two collected signals Q _{m (f,} j) and the A ^* is the complex conjugate of A, the E [A] as a function that takes an average of A,
In the input / output correlation coefficient calculation unit, the power spectrum P ₀₀ of the pseudo echo signal and the power spectrum P _mm of the m-th channel received signal are calculated.

As calculated, the pseudo echo signal and the cross spectrum P _{0m (f,} j) between the received signal of the m channels and the cross spectrum P _m between the received signal of the m 'channel and the reception signal of the m-channel _'m (f, j)

And a cross spectrum Q ₀ (f, j) between the pseudo echo signal and the second sound pickup signal and a cross spectrum Q _m (f) between the m-th channel received signal and the second sound pickup signal. , J)

As sought
In the input / output transfer characteristic estimation unit, the input / output transfer characteristic G (f, j) is

Estimated as
In the residual echo prediction unit, the residual echo component is

Predict as
And a multi-channel echo canceller. M speakers (M is an integer equal to or greater than 2) and N microphones (N is an integer equal to or greater than 1) are arranged in a common sound field. A multi-channel echo canceller that cancels echo that circulates into the microphone,
An echo prediction unit that generates the pseudo echo signal by inputting the received signal into the pseudo echo path by the adaptive filter;
A subtractor for subtracting the pseudo echo signal from a first sound pickup signal picked up by the microphone to obtain a second sound pickup signal;
An echo path estimator that updates a filter coefficient of the adaptive filter using the second collected sound signal and the received signal;
A frequency domain converter that converts the pseudo echo signal, the second sound collection signal, and the received signal into a frequency domain signal, respectively;
A first input / output correlation coefficient calculating unit that obtains a cross spectrum between each M-channel received signal and the second sound pickup signal and a power spectrum of each channel of the M-channel received signal;
First, an input / output transfer characteristic is estimated for each frequency using a cross spectrum between each M-channel received signal and the second sound pickup signal and a power spectrum of each channel of the M-channel received signal. 1 input / output transfer characteristic estimation unit;
A first residual echo prediction unit that predicts a first residual echo component included in the second collected sound signal from the estimated input / output transfer characteristic of the M channel reception signal in the frequency domain;
A subtractor for subtracting the predicted first residual echo component from the second collected sound signal in the frequency domain to obtain a third collected signal;
A second input / output correlation coefficient calculating unit for obtaining a power spectrum of the pseudo echo signal and a cross spectrum between the third sound pickup signal and the pseudo echo signal;
A second input / output transfer characteristic estimator for estimating an input / output transfer characteristic for each frequency using a cross spectrum between the power spectrum of the pseudo echo signal, a third sound pickup signal, and the pseudo echo signal;
A second residual echo prediction unit that predicts a second residual echo component included in the third sound pickup signal from the input / output transfer characteristic estimated as the pseudo echo signal in the frequency domain;
A subtractor for subtracting the predicted second residual echo component from the third collected signal in the frequency domain to obtain a transmission signal;
A time domain conversion unit for converting the transmission signal obtained by the subtraction unit into a time domain signal;
A multi-channel echo canceller.   A program for causing a computer to execute the multi-channel echo cancellation method according to any one of claims 1 to 5.

Images