JP5086969B2 - Echo canceling apparatus, method thereof, program thereof, and recording medium thereof - Google Patents

Description

  The present invention relates to an echo canceling apparatus, a method thereof, a program thereof, and a recording medium thereof used in, for example, a loud stereo communication conference.

FIG. 1 shows a partial block diagram of a functional configuration example of a conventional echo canceling apparatus. A conventional echo canceling device is described in Non-Patent Document 1, for example. 1 includes two adaptive filter units 7 ₁ and 7 ₂ , two pseudo echo path units 5 ₁ and 5 ₂ , an addition unit 8, and a subtraction unit 6. In this example, as an environment to which the echo cancellation apparatus is applied, a stereo signal, that is, received signals x ₁ (k) and x ₂ (k) (received from the receiving ends 1 ₁ and 1 ₂ ( However, k is a discrete time) is reproduced by the reproduction means (for example, speakers) 2 ₁ and 2 ₂ , and the collected sound signal collected by the sound collection means (for example, microphone) 3 is input to the transmitting end 4. The case where the environment is called will be described. In the following description, symbols indicating vectors or matrices are shown in bold, scalar quantities are shown in lower case, and in this [Background Art], the description of each component will be simplified.

Received signals x ₁ (k) and x ₂ (k) from the receiving ends 1 ₁ and 1 ₂ are respectively input to the adaptive filter units 7 ₁ and 7 ₂ and the pseudo-echo path units 5 ₁ and 5 ₂ . Each echo path unit ₅ 1, _{5 2,} adaptive filter vector _W 1 _{(k), W} 2 has a (k), the received signal _x 1 _(k), adapted respectively to _x 2 (k) filter vector _{W 1} By convolving (k) and W ₂ (k), y ₁ (k) and y ₂ (k) are generated. That is, the following expression is calculated.
y ₁ (k) = W ₁ ^T (k) X ₁ (k) (1)
y ₂ (k) = W ₂ ^T (k) X ₂ (k) (2)
However, W _m (k) = [w _m1 (k), w _m2 (k),. . . , W _mL (k)] (3)
X _m (k) = [x _m (k), x _m (k−1),. . . , X _m (k−L + 1)] (4)
Here, m is the number of received signals (number of channels), m = 1, 2 in this example, T indicates transposition of the vector, L indicates the filter length (number of taps) of the adaptive filter, and w _mj (K) (j = 1 to L) indicates each coefficient of the adaptive filter vector W _m (k).

Then, the sum of y ₁ (k) and y ₂ (k) is calculated by the adder 8, that is, y (k) = y ₁ (k) + y ₂ (k) is calculated, and the total simulated echo signal y (K) is required. The total simulated echo signal y (k) is input to the subtraction unit 6.

On the other hand, when the reproduced sound reproduced from the reproducing means 2 ₁ , 2 ₂ is collected by the sound collecting means 3, the collected sound signal is set as an echo signal d (k). The subtracting unit 6 obtains a residual signal e (k) by calculating a difference between the echo signal d (k) and the pseudo echo signal y (k). Specifically, it is obtained by the following formula.
e (k) = d (k) -y (k) (5)

The adaptive filter units 7 ₁ and 7 ₂ sequentially use the residual signal e (k) and the received signals x ₁ (k) and x ₂ (k) to sequentially apply adaptive filter vectors W ₁ (k) and W ₂ (k ). Then, the estimated adaptive filter vectors W ₁ (k) and W ₂ (k) are copied to the pseudo echo path sections 5 ₁ and 5 ₂ . The adaptive filter vectors W ₁ (k) and W ₂ (k) are updated sequentially as follows.
W ₁ (k + 1) = W ₁ (k) + μΔW ₁ (k) (6)
W ₂ (k + 1) = W ₂ (k) + μΔW ₂ (k) (7)

Here, μ is a parameter for controlling the update amount. ΔW _m (k) indicates an update amount at time k, and differs depending on the adaptation algorithm. For example, the learning identification method is as follows.

By the way, two speakers and two microphones are necessary to perform a communication conference using stereo signals. The voice spoken by the speaker is picked up by two microphones (as described above, only one microphone is shown in FIG. 1). At this time, since the voice is generated by convolution with each impulse response from the speaker to each microphone, the collected stereo signal has a strong cross-correlation and the cross-correlation does not change. The adaptive filter including the learning identification method updates (corrects) the adaptive filter vector so that the echo signal e (k) becomes zero. However, when H _m (m = 1 to M in this example, m = 1, 2) is defined as an impulse response from the speaker 2 _m to the microphone 3, the cross-correlation of the stereo signal is strong and the cross-correlation varies. If not, it is not a true solution with W _m (k) = H _m ,
d (k) -y (k)
= H ₁ ^TX ₁ (k) + H ₂ ^TX ₂ (k) − (W ₁ ^TX ₁ (k) + W ₂ ^TX ₂ (k))
= 0 (10)
To find a local solution. If such a local solution is obtained, an echo is generated when the cross-correlation of the stereo signal fluctuates, such as when the speaker moves or the speaker changes.

In order to avoid this problem, a stereo echo cancellation apparatus may add non-linear preprocessing to the received signal. Nonlinear preprocessing can change the cross-correlation of stereo signals. However, if the amount of nonlinear preprocessing added is large, the adaptive filter does not fall into the local solution, but the audio quality and stereo localization are degraded. Often. If the amount of non-linear processing to be applied is small, the adaptive filter does not fall into a local solution, and the speech quality and stereo localization are less degraded, but the convergence speed until the true value of the adaptive filter is obtained becomes slow. In other words, depending on the amount of non-linear preprocessing to be applied, there is a trade-off between the convergence speed until the true value of the adaptive filter is obtained and the improvement in reproduction sound quality and the stability of localization. As a technique for solving this problem, paying attention to the fact that there is no correlation in the sum / difference signal of the stereo signal, a method (for example, Patent Document 1) for performing a cross spectrum calculation of the sum / difference signal and the collected sound signal or an orthogonalization filter is used. Thus, a method has been proposed in which a stereo signal is decorrelated and the decorrelated stereo signal is reproduced from a speaker (Patent Document 2).
Akira Nakagawa and Yoichi Haneda “A Study on Stereo Echo Canceller Considering Power Difference between Stereo Signals”, IEICE, vol. 86, no. 10, pp. 989-997, Oct. 2003. Japanese Patent No. 3506138 Japanese Patent No. 3608525

In the invention described in Non-Patent Document 1, the amount of non-linear preprocessing of the stereo echo cancellation apparatus cannot be increased so much as to maintain the reproduced sound quality. As a result, it is difficult to increase the convergence speed to the true value where W _m (k) = H _m, and the time until the echo can be canceled becomes long.

  Further, since the invention described in Patent Document 1 assumes that the sum / difference signal is uncorrelated, there is a case where the filter coefficient cannot be estimated at all and the echo cannot be canceled if the uncorrelated assumption is not satisfied.

  Further, in the invention described in Patent Document 2, since the orthogonalized stereo signal is decorrelated every fixed frame, there is a problem that the localization feeling is likely to fluctuate in units of frames when reproduced as it is.

In the echo canceling apparatus of the present invention, M received signals from M (M ≧ 2) receiving ends are converted and emitted into reproduced sound by M reproducing means, and N (N ≧ 1) collected sound signals. An echo signal obtained by collecting the reproduction sound by the sound collecting means is removed from the sound collecting signal input to the means. The echo cancellation apparatus includes a vector orthogonalizing unit, a vector normalizing unit, MN pseudo echo path units, N adding units,
For M received signal vectors of the M received signals, a vector orthogonalizing unit for obtaining M orthogonal vectors orthogonal to each other, N subtracting units, and MN orthogonal normalizing adaptive units, Have. The vector normalization unit obtains M orthogonal normalization vectors by normalizing the M orthogonal vectors. Each of the MN pseudo echo path units obtains a pseudo echo signal by convolving an adaptive filter estimated to the received signal. Each of the N adders adds up the M pseudo echo signals to obtain a total pseudo echo signal. Each of the N subtracting units obtains a residual signal by subtracting the collected sound signal from the total simulated echo signal. Each of the MN orthogonal normalization adaptation units obtains an update amount of the adaptive filter from a residual signal, the M reception signals, and the M orthogonal normalization vectors , The vector orthogonalizing unit sets the m-th received signal vector to X _m (k) (m = 1,..., M, and so on), and sets the m-th orthogonal vector to Z _m (k). The M orthogonal vectors Z _m (k) are obtained in accordance with (11) , and the n (n = 1,..., N) th orthogonal normalization adaptive unit transmits the _mth received signal vector X _m ( The update amount of the adaptive filter for k) is ΔW _mn (k), S _m (k) is the mth orthogonal normalization vector, and e _n (k) is the error signal for the nth sound collecting means. (17) by the adaptive filter update amount ΔW _mn (k) Seek.

  According to the echo cancellation apparatus of the present invention, the cross-correlation can be weakened by obtaining M orthogonal vectors that are orthogonal to each other for the M received signal vectors, so that the convergence speed of the adaptive filter can be increased. it can.

  The best mode for carrying out the invention will be described below. In addition, the same number is attached | subjected to the process which performs the structure part which has the same function, and the same process, and duplication description is abbreviate | omitted.

FIG. 2 shows a functional configuration example of the echo cancellation apparatus 100 of the first embodiment, and FIG. 3 shows a main processing flow. In this embodiment, an environment to which the echo cancellation apparatus 100 is applied will be described. M (M ≧ 2) receiving ends 1 _m are respectively connected to the corresponding speakers 2 _m , and receiving signals x _m (k) from the M receiving ends (m = 1,..., M). Are converted into reproduced sound by the M reproducing means 2 _m and emitted. The reproduced sound is picked up by one sound pickup means 3 and converted into a sound pickup signal, and this sound pickup signal (that is, a signal input by the reproduction sound wrapping around the sound pickup means 3) is echo signal d. (K). The echo cancellation apparatus 100 includes a vector orthogonalization unit 8, a normalization unit 9, M mth orthogonal normalization adaptation units 10 _m (m = 1,..., M), and M mth pseudo echo path units. 5 _m , an addition unit 7, and a subtraction unit 6. The vector orthogonalizing unit 8 has M m-th orthogonalizing units, and the normalizing unit 9 has M m-th normalizing units. As described above, the received signal vector is set to X _m (k) (see Equation (4)). In the present invention, it is assumed that the cross-correlation is changed to some extent. When this echo canceling device is used in a teleconference device, the causes of fluctuations in cross-correlation are speech encoding on the transmitting side, speech decoding on the receiving side, and transmission noise suppression processing on the transmitting side. This is a case in which reception noise suppression processing on the reception side, noise mixing in each microphone on the transmission side, and other nonlinear processing on the transmission side / reception side are performed. In general, a signal input to a microphone is generally subjected to non-linear processing such as noise suppression, and it is assumed that fluctuations in cross-correlation required by the present invention inevitably occur.

First, when the M received signals x _{m (k)} is input to the m orthogonal unit 8 _m, respectively, the first m orthogonalizing means 8 _m for the M received signal vector of the M received signals, M orthogonal vectors Z _m (k) that are orthogonal to each other are obtained (step S2). That is, it is only necessary to obtain an orthogonal vector such that the inner product value of all vectors is zero. As a method of obtaining the M orthogonal vectors Z _m (k), for example, it can be obtained as follows.

In this example, the first orthogonalizing means 8 ₁ ,. . . , The m orthogonal unit ₈ m,. . . ,
The M-th orthogonalizing means 8 _M is respectively Z ₁ (k),. . . , Z _m (k),. . . , Z _M (k). In the present invention, since it is empirically known that the cross-correlation is weakened by obtaining the orthogonal vector for the received signal vector X _m (k), the convergence speed of the adaptive filter can be increased. The obtained orthogonal vectors Z _m (k) (m = 1,..., M) are respectively input to the corresponding m-th normalization means 9 _m .

The m-th normalization means 9 _m obtains an orthogonal normalized vector S _m (k) by normalizing the orthogonal vector Z _m (k) (step S4). That is, the normalization unit 9 obtains M orthogonal normalization vectors by normalizing M orthogonal vectors, respectively (step S4). For example, normalization can be performed as follows.

By performing such normalization, the convergence speed of the adaptive filter can be increased even when the power difference between the received signals is large. Orthogonal normalized vector than the m normalization means _{9 m S m (k) (} m = 1, ..., M) are all input to each of the m orthogonal normalization adaptation section 10 _m.

There are M orthogonal normalization adaptation units, and each orthogonal normalization adaptation unit has a residual signal e (k) from the subtraction unit 6, M received signal vectors X _m (k), and M orthogonal normalization units. The update amount ΔW _m (k) of the adaptive filter vector W _m (k) is obtained from the normalized vector S _m (k) (step S12). For example, as a first method, the update amount ΔW _m (k) can be obtained by the following equation.

Then, the m-th orthogonal normalization adaptation unit 10 _m sets the filter coefficient W (k) of the m-th pseudo echo path unit to W _m (k + 1) = W _m (k) + μΔW _m (k) (14)
Update with

In addition, either the denominator on the right side of Equation (13) or the numerator computation processing may be performed first. In addition, ΔW _m (k) (m = 1,..., M) has a common denominator in all arithmetic processes. Therefore, the denominator may be calculated first and stored in a storage unit (not shown), and then the update amount ΔW _m (k) of the adaptive filter vector W _m (k) may be obtained.

Next, as a second method for obtaining the update amount ΔW _m (k), for example, the following update formula using p past values may be used.
ΔW _m (k) = U _m (k) [Σ _{i = 1} ^m U _i ^T (k) V _i (k)] ⁻¹ E (k)
U _m (k) = [S _m (k), S _m (k−1),. . . , S _m (k−p + 1)]
V _m (k) = [X _m (k), X _m (k−1),. . . , X _m (k−p + 1)]
E (k) = [e (k), (1-μ) e (k−1),. . . , (1-μ) ^p−1 e (k−p + 1)]

The processing contents of the m-th pseudo echo path section 5 _m , the adding section 7 and the subtracting section 6 are the same as those in the prior art, but will be briefly described just in case. The m-th pseudo echo path unit 5 _m obtains the m-th pseudo echo signal y _m (k) by convolving W _m (k) with the received signal vector X _m (k) (step S6). That is, it calculates | requires with the following formula | equation.
y _m (k) = W _m (k) X _m (k) (15)

The adder 7 obtains the total simulated echo signal y (k) by adding M pseudo echo signals y _m (k). That is, it calculates | requires by the following formula | equation (step S8).
y (k) = Σ _{m = 1} ^M y _m (k) (16)

The subtracting unit 6 obtains a residual signal e (k) by subtracting the collected sound signal d (k) from the total simulated echo signal y (k) (step S10). Then, the residual signal e (k) is input to each of the mth orthogonal normalization adaptation unit 10 _m .

  In the above description and FIG. 2, it has been described that the vector orthogonalizing unit 8 and the normalizing unit 9 perform individual processing. However, the vector orthogonalizing unit 8 and the normalizing unit 9 are integrated. Also good. Further, an echo canceling device including at least one of the reproducing means and the sound collecting means may be used.

  Thus, in an adaptive filter of a multi-channel (two channels in stereo) echo cancellation device, it is necessary to reduce non-linear pre-processing for changing the cross-correlation of the collected sound signals in order to ensure reproduction sound quality and a sense of localization There is. Therefore, it takes time for the adaptive filter to converge to the true value. In this embodiment, it has been empirically found that cross-correlation can be weakened without deteriorating sound quality and localization by orthogonalizing the received signal whose cross-correlation has been fluctuated to some extent before convolution with the adaptive filter. The convergence speed can be increased. In addition, by normalizing the received signal after orthogonalization, the convergence speed can be increased regardless of the power difference of the original received signal.

In the first embodiment, the case where there is one sound collecting unit has been described. In the second embodiment, a case will be described in which there are N sound collecting means (N ≧ 2, and hereinafter referred to as sound collecting means 3 _n (n = 1,..., N)). FIG. 4 shows an echo cancellation apparatus 200 according to the second embodiment. For simplification of description, the vector orthogonalization unit 8 and the normalization unit 9 are collectively referred to as a first calculation unit 20, and M m-th orthogonal normalization adaptation units 10 _m and M m-th pseudo echoes are included. The path unit 5 _m , the adding unit 7 and the subtracting unit 6 are collectively referred to as a second computing unit 30 _n (n = 1,..., N). The second calculation unit 30 _n is connected to the corresponding sound collecting means 3 _n . That is, there are MN orthogonal normalization adaptation units and pseudo echo path units, respectively, which will be described below as mn orthogonal normalization adaptation units 10 _mn (m = 1,..., M n = 1,. N), the mnth pseudo echo path section 5 _mn .

Orthogonal normalization vectors S _m from the m-th normalization unit 9 _m (m = 1,..., M) in the first calculation unit 20 are respectively represented by the mn orthogonal normalization adaptation unit 10 to which the subscript m corresponds. Input to _mn .

The update amount obtained by the mn orthogonal normalization adaptation unit 10 _mn is ΔW _mn (k), and the error signal output from the subtraction unit 6 _n in the second calculation unit 30 _n is e _n (k). Let W _mn (k) be the adaptive filter vector of the mn pseudo echo path section. The mn orthogonal normalization adaptation unit 10 _mn can obtain the update amount ΔW _mn (k) as follows, for example.

Then, the mn orthogonal normalization adaptation unit 10 _mn updates the adaptive filter vector W _mn (k) included in the mn pseudo echo path unit 5 _mn as follows.
W _mn (k + 1) = W _mn (k) + ΔW _mn (k) (18)
Thus, even if there are two or more sound collecting means, the echo canceling apparatus of the present invention can be effectively applied.

[simulation result]
Next, simulation results for confirming the effects of the conventional echo canceling apparatus and the present application will be shown. The simulation conditions are described. Two reproducing means (that is, a stereo system) were used, and the received signal from the receiving end 1 _m was convoluted with two impulse responses measured on white noise to obtain a stereo signal. And in the conventional echo cancellation apparatus, the half wave rectification addition shown below was performed as nonlinear pre-processing. However, r = 0.3.

  The algorithm used in the conventional echo cancellation apparatus is a learning identification method. Two independent white noises were exponentially attenuated to obtain a near-end impulse response. The experimental results are shown in FIG. The horizontal axis in FIG. 5 represents time, and the vertical axis represents coefficient error. The equation of the coefficient error ce (k) is expressed by the following equation:

  In this simulation, the convergence time until the coefficient error becomes −20 dB is compared (portion indicated by a broken line in FIG. 5). As understood from FIG. 5, the method of the present invention can increase the convergence speed by about 1.8 times compared with the conventional learning identification method.

FIG. 6 shows a simulation result for confirming the effect when there is a power difference between the two received signals x ₁ (k) and x ₂ (k). 6 differs from the simulation conditions in the case of FIG. 6 in that the method of generating the received signal is different, and between the stereo signals (that is, between the received signals x ₁ (k) and x ₂ (k)) in the response measurement of the impulse response. Measurements were taken in an arrangement that would cause a power difference. Other simulation conditions are the same as in FIG. As can be understood from FIG. 6, the method of the present invention can increase the convergence speed by about 2.5 times as compared with the conventional learning identification method.

  In addition to the above embodiments, the echo canceling apparatus according to the present invention is not limited to the above-described embodiments, and can be appropriately changed without departing from the spirit of the present invention. Further, the processes described in the echo canceling apparatus are not only executed in time series in the order described, but may be executed in parallel or individually as required by the processing capability of the apparatus that executes the processes. .

  Further, when the processing in the echo cancellation apparatus of the present invention is realized by a computer, the processing contents of the functions that the echo cancellation apparatus should have are described by a program. Then, by executing this program on the computer, the processing function in the echo canceling apparatus is realized on the computer.

  The program describing the processing contents can be recorded on a computer-readable recording medium. As the computer-readable recording medium, for example, any recording medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, and a semiconductor memory may be used. Specifically, for example, as a magnetic recording device, a hard disk device, a flexible disk, a magnetic tape, or the like, and an optical disk such as a DVD (Digital Versatile Disc), a DVD-RAM (Random Access Memory), a CD-ROM (Compact Disc Read Only Memory), a CD-R (Recordable). ) / RW (ReWritable), etc., magneto-optical recording medium, MO (Magneto-Optical disc), etc., semiconductor memory, EEP-ROM (Electronically Erasable Programmable-Read Only Memory), etc. can be used.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Furthermore, the program may be distributed by storing the program in a storage device of the server computer and transferring the program from the server computer to another computer via a network.

  A computer that executes such a program first stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its own storage device. When executing the process, the computer reads a program stored in its own recording medium and executes a process according to the read program. As another execution form of the program, the computer may directly read the program from a portable recording medium and execute processing according to the program, and the program is transferred from the server computer to the computer. Each time, the processing according to the received program may be executed sequentially. Further, the above-described processing may be executed by a so-called ASP (Application Service Provider) type service that realizes a processing function only by an execution instruction and result acquisition without transferring a program from the server computer to the computer. Good. Note that the program in this embodiment includes information that is used for processing by an electronic computer and that conforms to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer).

  In this embodiment, the echo canceling apparatus is configured by executing a predetermined program on a computer. However, at least a part of these processing contents may be realized by hardware.

The block diagram which shows the function structural example of the conventional echo cancellation apparatus. FIG. 3 is a block diagram illustrating a functional configuration example of the echo cancellation apparatus according to the first embodiment. FIG. 3 is a diagram illustrating a main processing flow of the echo cancellation apparatus according to the first embodiment. FIG. 6 is a block diagram illustrating an example of a functional configuration of an echo cancellation apparatus according to a second embodiment. The figure which shows a simulation result. The figure which shows the simulation result at the time of changing conditions.

Claims (4)

M received signals from M (M ≧ 2) receiving ends are converted and emitted into reproduced sound by M reproducing means, and are input to N (N ≧ 1) sound collecting means. From the echo canceller for removing the echo signal from which the reproduced sound is picked up by the sound pickup means,
A vector orthogonalization unit for obtaining M orthogonal vectors orthogonal to each other for M reception signal vectors of the M reception signals;
A vector normalization unit for obtaining M orthogonal normalization vectors by normalizing the M orthogonal vectors;
Each of the MN pseudo-resonance path units for obtaining the pseudo-resonance signal by convolving the estimated adaptive filter with the received signal,
N adders for obtaining a total simulated echo signal by adding M pseudo echo signals,
N subtracting units each for obtaining a residual signal by subtracting the collected sound signal from the total simulated echo signal;
Each comprising MN orthogonal normalization adaptation units for obtaining an update amount of the adaptive filter from a residual signal, the M received signal vectors, and the M orthogonal normalization vectors ,
The vector orthogonalization unit includes:
If the m-th received signal vector is X _m (k) (m = 1,..., M, and so on), and the m-th orthogonal vector is Z _m (k) , respectively .

To obtain the M orthogonal vectors Z _m (k),
The n (n = 1,..., N) th orthogonal normalization adaptation unit sets ΔW _mn (k) as the adaptive filter update amount for the _mth received signal vector X _m (k) , and S _{If m} (k) is the mth orthogonal normalization vector and e _n (k) is the error signal for the nth sound collection means,

An echo cancellation apparatus characterized by obtaining an update amount ΔW _mn (k) of the adaptive filter by M received signals from M (M ≧ 2) receiving ends are converted and emitted into reproduced sound by M reproducing means, and are input to N (N ≧ 1) sound collecting means. From the echo cancellation method for removing the echo signal from which the reproduced sound is picked up by the sound pickup means,
A vector orthogonalization process for obtaining M orthogonal vectors orthogonal to each other for M received signal vectors of the M received signals;
A vector normalization process for obtaining M orthogonal normalization vectors by normalizing the M orthogonal vectors;
By convolving the estimated adaptive filter with the received signal, a pseudo echo path process for obtaining a pseudo echo signal,
N addition processes for obtaining a total pseudo echo signal by adding the M pseudo echo signals;
N subtraction processes for obtaining a residual signal by subtracting the collected sound signal from the total simulated echo signal;
Possess a residual signal, wherein the M pieces of received signals, wherein the M pieces orthogonal normalized vector of the MN number of orthogonal normalized adaptation process for obtaining the update amount of the adaptive filter from a
The vector orthogonalization process is:
If the m-th received signal vector is X _m (k) (m = 1,..., M, and so on), and the m-th orthogonal vector is Z _m (k) , respectively .

To obtain the M orthogonal vectors Z _m (k),
In the orthogonal normalization adaptation process, the update amount of the adaptive filter for the m-th received signal vector X _m (k) is ΔW _mn (k), S _m (k) is the m-th orthogonal normalization vector, If e _n (k) is an error signal for the nth sound collecting means,

An echo cancellation method characterized in that the update amount ΔW _mn (k) of the adaptive filter is obtained by A program for operating a computer as the echo canceling device according to claim 1 . A computer-readable recording medium on which the program according to claim 3 is recorded.

Images