JP6180689B1 - Echo canceller apparatus, echo cancellation method, and echo cancellation program - Google Patents

Abstract

The echo canceller apparatus (100) includes an adaptive filter unit (101) that removes a pseudo echo component from a microphone input signal (y (n)) and outputs a residual signal (r (n)), a microphone input signal, and a residual signal. A pERL calculation unit (106) for calculating a pERL value indicating the ratio of the difference signal, an echo signal (d (n)) of the microphone input signal, and a residual echo signal obtained by subtracting the pseudo echo component from the echo signal An ERLE calculation unit (105) that calculates an ERLE value indicating a ratio to (e (n)), a pERL decrease degree calculation unit (107) that calculates a decrease degree indicating a difference between the ERLE value and the pERL value, and an ERLE A suppression amount calculation unit (108) for calculating a residual echo suppression amount (γ (ω)) from the value (t) and the reduction degree (k), and multiplying the residual signal by the residual echo suppression amount, an output signal Produce residual residue A co-suppression processing unit (109).

Description

  The present invention relates to an echo canceller apparatus, an echo cancellation method, and an echo cancellation program used in a voice communication system.

  The echo canceller apparatus is used in an audio communication system such as a telephone communication system and a video conference system in order to cancel an echo signal based on sound (echo) that circulates from a speaker to a microphone. In general, in an echo canceller, an adaptive filter estimates (learns) an echo path from an echo signal and a received signal, generates an echo replica that is a pseudo echo signal based on the estimation result, and an echo from the transmitted signal by a subtractor. Echo cancellation processing is performed by reducing replicas. Many echo canceller apparatuses include a residual echo suppression processing unit that suppresses residual echo signals in order to reduce residual echo signals remaining in the transmission signal after echo cancellation processing using an adaptive filter.

  However, in the conventional echo canceller, when the microphone input signal is a signal obtained by superimposing the echo signal and the near-end signal (a signal based on background noise, near-end speaker voice, etc.), the echo signal is accurately erased. Difficult to do. Particularly, in the case of double talk (DT: Double Talk) in which the near-end speaker sound and the echo (far-end speaker sound) output from the speaker are simultaneously input to the microphone, the echo path is estimated (learning) by the adaptive filter. Errors tend to occur, and the residual echo amount estimation accuracy in the residual echo suppression processing unit is reduced. Further, at the time of DT, the difference between the echo signal and the echo replica becomes large, and the amount of the residual echo signal remaining in the transmission signal after the echo cancellation processing increases.

  As these countermeasures, Patent Document 1 discloses an echo path coupling amount obtained by dividing the echo signal power by the received signal power (that is, an acoustic coupling amount obtained as a power ratio between the echo signal and the received signal). The method of estimating the echo signal power by multiplying the received signal power and suppressing the echo signal using the estimation result is proposed.

Japanese Patent No. 3420705 (paragraphs 0089 to 0107, FIG. 1)

Jacob Benesty, Dennis R. Morgan, and Jun. H. Cho, “A New Class of Doubletalk Detectors Based on Cross-Correlation” IEEE Transactions on Speech and Audio Processing, Vol.8, No.2, pp.168-172 , Mar. 2000. Asano Takuma, Acoustic Technology Series 16, “Array signal processing for acoustics, -Localization, tracking and separation of sound sources”, Chapter 3, 58-59 Page, February 25, 2011, issued by Corona Inc.

  However, in the echo canceller described in Patent Document 1, since the echo path coupling amount, which is the power ratio between the echo signal and the received signal, is used for calculating the power of the residual echo signal after the echo cancellation processing, The estimation accuracy of the echo signal is low. When the residual echo amount estimation accuracy is low in the residual echo suppression processing unit, the near-end signal (including the near-end speaker voice signal) of the transmitted signal is suppressed or the residual echo signal is sufficiently reduced. There is a problem that can not be.

  An object of the present invention is to provide an echo canceller apparatus, an echo cancellation method, and an echo cancellation program that can suppress a residual echo signal without suppressing a near-end signal.

An echo canceller apparatus according to an aspect of the present invention is an echo canceller apparatus used in an audio communication system having a microphone and a speaker, and an echo that removes a pseudo echo component from a microphone input signal and outputs a residual signal An erasure unit; a pERL calculation unit that calculates a pERL value indicating a ratio between the microphone input signal and the residual signal; an echo signal based on an echo input from the speaker to the microphone of the microphone input signal; An ERLE calculation unit for calculating an ERLE value indicating a ratio of the echo signal to the residual echo signal obtained by subtracting the pseudo echo component, and a pERL decrease for calculating a reduction degree indicating a difference between the ERLE value and the pERL value a degree calculating unit, a value showing the decrease level in the linear value is K, the value showing the ERLE value in a linear value is T and A, by multiplying the suppression amount calculating unit for calculating a residual echo suppression amount from the formula (K-1) T / K (T-1), the residual echo suppression amount to the residual signal to produce an output signal And a residual echo suppression processing unit.

  According to the present invention, it is possible to suppress the residual echo signal without suppressing the near-end signal.

It is a block diagram which shows schematic structure of the echo canceller apparatus which concerns on Embodiment 1 of this invention. It is a figure which shows the graph for demonstrating that a pERL value reduces, so that the value of a near end signal becomes large. It is a figure which shows the graph of the pERL value which decreases monotonically, so that the value of a near end signal becomes large. FIG. 4 is a graph showing the pERL value t [dB] when the adaptive filter is stabilized and k [dB] which is the degradation amount of the pERL value due to the occurrence of double talk in the graph of the pERL value shown in FIG. 3. is there. 1 is a hardware configuration diagram illustrating an echo canceller apparatus according to a first embodiment. FIG. 5 is another hardware configuration diagram illustrating the echo canceller apparatus according to the first embodiment. 5 is a flowchart showing an operation (echo cancellation method) of the echo canceller apparatus according to the first embodiment. It is a block diagram which shows schematic structure of the echo canceller apparatus which concerns on Embodiment 2 of this invention. 6 is a flowchart showing an operation (echo cancellation method) of the echo canceller apparatus according to the second embodiment.

<< 1 >> Embodiment 1
<< 1-1 >> Configuration FIG. 1 is a block diagram showing a schematic configuration of an echo canceller apparatus 100 according to Embodiment 1 of the present invention. The echo canceller apparatus 100 is an apparatus that can execute the echo cancellation method according to the first embodiment. The echo canceller apparatus 100 can constitute a part of the voice communication system 10 having the microphone 11 and the speaker 12. The voice communication system 10 is communicably connected to another voice communication system (not shown) via a signal transmission line such as a telephone line and the Internet. Examples of the voice communication system 10 include a telephone communication system (speaking call device) having a hands-free call function and a video conference system (video conference device).

  As shown in FIG. 1, the echo canceller apparatus 100 includes an adaptive filter 101a that performs estimation (learning) of an echo path. The adaptive filter 101a updates (calculates) the filter coefficient to estimate the echo path, and based on the echo path estimation result, the pseudo echo resembling the echo signal based on the echo input from the speaker 12 to the microphone 11 An echo replica that is a signal is generated. The echo canceller apparatus 100 includes a subtractor 101b that subtracts an echo replica from a microphone input signal y (n). The adaptive filter 101a and the subtractor 101b constitute an echo cancellation unit (adaptive filter unit) 101.

  The echo canceller apparatus 100 preferably includes a double talk determiner (DT determiner) 102 that determines whether the sound input to the microphone 11 is in a double talk (DT) state. As the DT determiner 102, for example, the one described in Non-Patent Document 1 can be used. However, the DT determiner 102 may be another DT determiner as long as it has a function of determining whether or not it is in the DT state (that is, the presence or absence of DT). The DT determiner 102 controls the adaptive filter 101a so that the adaptive filter 101a does not estimate (learn) the echo path during DT.

The time is expressed as n, the echo signal based on the echo at time n is expressed as d (n), and the near-end signal indicating the sum of the near-end speaker voice and noise at time n is expressed as s (n). When the microphone input signal based on the sound input to the microphone 11 at time n is expressed as y (n), the following equation is established.
y (n) = d (n) + s (n)

を推定するためにフィルタ係数

を更新し、同時にエコーレプリカ

を生成する。The adaptive filter 101a has an echo path

Filter coefficients to estimate

Update and echo replica at the same time

Is generated.

  As the adaptive filter 101a, for example, a filter using a Normalized Least Mean Square (NLMS) algorithm can be employed. As a coefficient updating method using NLMS, for example, the method described in Non-Patent Document 2 can be used. The update method of NLMS is shown in (Formula 1) and (Formula 2).

（Ｎは、フィルタ長）と表される。上付きの「Ｔ」は、転置を意味する。In (Expression 1) and (Expression 2), r (n) represents a residual signal obtained by subtracting an echo replica from the microphone input signal y (n). In (Expression 1) and (Expression 2), the time series of the past N received signals x (n) is a vector expression composed of N elements,

(N is a filter length). The superscript “T” means transposition.

は、受話信号

と、フィルタ係数

との内積で算出され、

で表わされる。Also, the echo replica that is the second term on the right side of (Equation 1)

Is the incoming signal

And the filter coefficient

And the inner product of

It is represented by

である。In the denominator of the second term on the right side of (Equation 2),

It is.

  A fast Fourier transform (FFT) processing unit 103 converts a time domain microphone input signal y (n) and a time domain residual signal r (n) into a frequency domain microphone input signal Y (ω) and a frequency. Each is converted into a residual signal R (ω) of the region. Here, ω represents a frequency.

  The suppression amount estimation unit 104 receives the frequency domain microphone input signal Y (ω), the frequency domain residual signal R (ω), and the DT determination result. The suppression amount estimation unit 104 estimates (calculates) the residual echo suppression amount γ (ω) from the frequency domain microphone input signal Y (ω) and the frequency domain residual signal R (ω). The suppression amount estimation unit 104 is a characteristic part of the present invention.

  First, in the suppression amount estimation unit 104, pERL (n), which is a pERL value (pseudo echo return loss) that simulates the echo cancellation amount, is obtained from the microphone input signal y (n) and the residual signal r (n). ) And is defined as (Equation 3).

  Further, ERLE (n) which is an ERLE value (Echo Return Loss Enhancement) which is a true echo cancellation amount is expressed by (Equation 4) using an echo signal d (n) and a residual echo signal e (n). Defined in

であり、ＤＴ時（遠端音声であるエコーと近端音声とがマイク１１に入力している最中）には、正確に観測できない信号である。Here, the residual echo signal e (n) is

This is a signal that cannot be accurately observed during DT (while echoes that are far-end speech and near-end speech are being input to the microphone 11).

Further, the power average pERL _ave (n) with respect to the pERL value is defined as in (Expression 5).

In (Formula 5), E [•] is an operator representing a time average of values in parentheses. Further, in (Expression 5), if it is assumed that the contents (true number) of the common logarithm (log ₁₀ ) are uncorrelated between the near-end signal s (n) and the echo signal d (n), ).

In Equation 6, the {s (n)} ² times the average a is E [{s (n)} 2] is increased, it is known that pERL value decreases. This can be easily understood by graphing (Equation 6).

とすることによって得られる、以下のｆ（ｘ）を検討する。

In (Formula 6),

The following f (x) obtained by

FIG. 2 is a diagram illustrating a graph for explaining that the pERL value decreases as the value of the near- end signal increases. FIG. 2 shows a graph of f (x) where a = 15 and b = 2000. FIG. 2 shows that f (x) is a function that monotonously decreases with f (0) = b / a as the maximum value.

FIG. 3 is a graph showing a pERL value that monotonously decreases as the value of the near- end signal increases. The graph shown in FIG. 3 represents f (x) shown in FIG. 2, and the numerical value on the vertical axis is expressed in decibels (dB). From FIG. 3, it can be seen that in the data during an actual hands-free call, the echo cancellation amount drops by about 10 dB during double talk (that is, drops from about 23 dB to about 13 dB).

  Next, an ideal residual echo suppression amount γ (ω) in the residual echo suppression processing unit 109 will be described. The microphone input signal of frequency ω is represented as Y (ω), the near-end signal is represented as S (ω), the echo signal is represented as D (ω), the residual signal is represented as R (ω), and the residual echo signal is represented as E (ω). . An ideal residual echo suppression amount γ (ω) with respect to the frequency ω is expressed by (Expression 7).

  The residual signal R (ω) is an observable signal, but the near-end signal S (ω) is a signal that cannot be observed. Therefore, in the first embodiment of the present invention, the near-end signal S (ω) is estimated with high accuracy. The method will be described below.

  If near-end signals are not mixed, the pERL value matches the ERLE value, and if the learning of the adaptive filter 101a sufficiently proceeds, the pERL value becomes stable (for example, 30 dB) at each time. Using this property, the near-end signal power is estimated from the degradation amount of the pERL value (for example, 10 dB drop) when double talk occurs.

とし、適応フィルタ１０１ａが安定した際のＥＲＬＥ値をｔ［ｄＢ］とする。

And the ERLE value when the adaptive filter 101a is stabilized is t [dB].

  For example, the ERLE calculation unit 105 determines the presence or absence of the near-end signal s (n) based on the determination result of the DT determiner 102 (that is, there is no DT if there is no near-end signal), and the near-end signal s (n) If not, the ERLE value is calculated. The calculation of the ERLE value may be any of an average of several frames, a long-time moving average, an immediately preceding instantaneous value, and the like.

  The pERL calculation unit 106 calculates a pERL value, and the pERL decrease degree calculation unit 107 calculates a difference between the ERLE value and the pERL value. When formulating that the pERL value has deteriorated by k [dB] due to the occurrence of double talk, the following (Equation 9) is obtained.

  FIG. 4 is a diagram showing t [dB] which is an ERLE value when the adaptive filter 101a is stabilized on the graph of FIG. 3 and k [dB] which shows a deterioration amount of the pERL value due to the occurrence of double talk. is there. t [dB] and k [dB] are observable values.

とすると、（式９）における最下段の式は、下記の（式１０）のように変形できる。

Then, the lowermost expression in (Expression 9) can be transformed as shown in (Expression 10) below.

(Equation 10) can be transformed into the following (Equation 11), and x can be obtained from (Equation 11).

From these results, the ratio of a, b, and x can be obtained by the following (formula 12).

From the above, the residual echo suppression amount γ (ω) can be obtained from the following (formula 13).

  For example, the suppression amount calculation unit 108 calculates the residual echo suppression amount γ (ω) using (Equation 13).

を得る。The residual echo suppression processing unit 109 multiplies the microphone input signal Y (ω) by the residual echo suppression amount γ (ω) obtained by the suppression amount calculation unit 108 and outputs an output signal.

Get.

を生成する。Finally, a fast inverse Fourier transform (IFFT) processing unit 110 performs IFFT processing on the output signal of the residual echo suppression processing unit 109, and outputs the time domain output signal.

Is generated.

<< 1-2 >> Hardware Configuration The echo canceller apparatus 100 according to the first embodiment can be realized by hardware (H / W), software (S / W), or a combination of H / W and S / W. is there.

  FIG. 5 is a hardware configuration diagram illustrating the echo canceller apparatus 100 according to the first embodiment. FIG. 5 shows an example in which the echo canceller apparatus 100 is configured with H / W. In this case, the echo canceller apparatus 100 includes a processing circuit 81 configured by a semiconductor integrated circuit or the like, and an external storage device 82 that stores information used by the processing circuit 81 for processing for echo cancellation. The echo cancellation method executed by the processing circuit 81 is shown in FIG. The external storage device 82 is, for example, an HDD (hard disk drive) or an SSD (solid state drive) connected to the echo canceller device 100 directly or via a network. The echo canceller apparatus 100 includes a microphone 11 as a voice input unit for inputting near-end speaker voice, a speaker 12 as a voice output unit for outputting far-end speaker voice based on a received signal, and a signal transmission path. The communication unit 13 that communicates with another voice communication system is connected to the communication unit 13. The components 101 to 110 shown in FIG. 1 can be realized by the processing circuit 81 and the external storage device 82 shown in FIG.

  FIG. 6 is another hardware configuration diagram showing the echo canceller apparatus 100 according to the first embodiment. FIG. 6 shows an example in which the echo canceller apparatus 100 is configured with S / W. In this case, the echo canceller apparatus 100 includes a memory 92 as a storage device that stores a program (echo cancellation program according to the first embodiment) as software, and an information processing unit that executes the program stored in the memory 92. It can be realized using the processor 91 (for example, by a computer). The processing executed by the echo cancellation program is shown in FIG. In addition, the echo canceller apparatus 100 includes an external storage device 82 that stores information used for processing for echo cancellation. In this case, the components 101 to 110 shown in FIG. 1 are realized by the processor 91 and the memory 92 shown in FIG. A part of the echo canceller apparatus 100 shown in FIG. 1 is realized by the memory 92 shown in FIG. 6 and a processor 91 that executes a program, and the remaining part is realized by the processing circuit 81 shown in FIG. May be.

<< 1-3 >> Operation FIG. 7 is a flowchart showing the operation of the echo canceller apparatus 100 according to the first embodiment (echo cancellation method according to the first embodiment).

  First, the DT determiner 102 determines whether or not DT has occurred in the echo canceller apparatus 100, that is, the presence or absence of DT (step S1).

  If there is no DT (NO in step S1), the adaptive filter 101a performs coefficient update processing in the adaptive filter 101a (step S2), generates an echo replica, and generates a echo replica from the microphone input signal y (n). A process of generating the residual signal r (n) by reducing the echo replica is performed (step S3).

  If DT is present (YES in step S1), the adaptive filter 101a generates an echo replica without performing coefficient update processing in the adaptive filter 101a, and the subtractor 101b generates an echo replica from the microphone input signal y (n). To generate a residual signal r (n) (step S3).

  Next, the FFT processing unit 103 converts the microphone input signal y (n) and the residual signal r (n) into a frequency domain microphone input signal Y (ω) and a residual signal R (ω), respectively (step S4). ).

  Next, the pERL calculation unit 106 calculates a pERL value from the microphone input signal Y (ω) and the residual signal R (ω) in the frequency domain (step S5).

  Next, if there is no DT (NO in step S6), the ERLE calculation unit 105 calculates an ERLE value (step S7), and the pERL decrease degree calculation unit 107 calculates the decrease degree from the pERL value and the ERLE value. Calculate (step S8).

  If there is DT (YES in step S6), the ERLE calculation unit 105 does not calculate the ERLE value, and the pERL decrease degree calculation unit 107 stores the deterioration amount of the pERL value at the time of double talk held in advance ( For example, 10 dB is assumed to be the decrease degree of the pERL value (step S8).

  Next, the suppression amount calculation unit 108 calculates the residual echo suppression amount γ (ω) from the ERLE value and the degree of decrease (step S9).

  Next, the residual echo suppression processing unit 109 performs residual echo signal suppression processing by multiplying the residual signal R (ω) by the residual echo suppression amount γ (ω) (step S10).

  Next, IFFT processing section 110 transmits the output signal by performing a fast inverse Fourier transform on the output signal of residual echo suppression processing section 109.

<< 1-4 >> Effect As described above, in the echo canceller apparatus 100 and the echo cancellation method according to Embodiment 1, as shown in (Equation 13), the ERLE value when the adaptive filter 101a is stabilized The residual echo suppression amount γ (ω) is calculated using T (≡10 ^{t / 10} ) based on t [dB], and K (≡10 ^{k / 10} ) based on k [dB] which is the degree of decrease in the pERL value. Calculated. That is, in the first embodiment, by observing the echo cancellation amount of the adaptive filter 101a, the ratio of the near-end signal S (ω) in the residual signal R (ω), that is, (Equation 7) is obtained. It is possible to estimate the ideal residual echo suppression amount shown. For this reason, according to the echo canceller apparatus 100 and the echo cancellation method according to the first embodiment, high echo cancellation (suppression) performance can be exhibited without suppressing the near- end signal.

<< 2 >> Embodiment 2
FIG. 8 is a block diagram showing a schematic configuration of an echo canceller apparatus 200 according to Embodiment 2 of the present invention. 8, components that are the same as or correspond to the components shown in FIG. 1 are given the same reference numerals as those shown in FIG. The echo canceller apparatus 200 is an apparatus that can execute the echo cancellation method according to the second embodiment. The echo canceller apparatus 200 constitutes a part of the voice communication system 20 having the microphone 11 and the speaker 12. The voice communication system 20 is communicably connected to another voice communication system (not shown) via a signal transmission line such as a telephone line and the Internet.

  The echo canceller apparatus 200 according to the second embodiment is such that the suppression amount estimation unit 201 includes a noise level calculation unit 202 that calculates the level of noise input to the microphone 11 before the ERLE calculation unit 105. This is different from the echo canceller apparatus 100 according to the first embodiment.

  FIG. 9 is a flowchart showing the operation of the echo canceller apparatus 200 according to the second embodiment. 9, processing steps that are the same as or correspond to the processing steps shown in FIG. 7 are assigned the same reference numerals as those shown in FIG. Echo canceller apparatus 200 differs from echo canceller apparatus 100 according to the first embodiment shown in FIG. 7 in that it has step S21 for calculating the noise level after step S5.

  The level of noise input to the microphone 11 can be calculated by observing the residual signal R (ω) and calculating, for example, by averaging the minimum frame power of the section. By subtracting the calculated noise level from the residual signal R (ω), the ERLE value can be calculated more accurately. As the accuracy of the ERLE value increases, the accuracy of the suppression amount calculation unit 108 increases as a result, and the echo cancellation performance improves.

  Note that the hardware configuration shown in FIGS. 5 and 6 can also be applied to the second embodiment.

  Regarding points other than those described above, the second embodiment is the same as the first embodiment.

  The present invention provides a conversation between a far-end speaker and a near-end speaker using a communication system, such as a telephone (including an in-vehicle telephone) for performing a hands-free call, a telephone communication system including a telephone, and a video conference system. It can be applied to all systems having a function to

  10, 20 Audio communication system, 11 Microphone (audio input unit), 12 Speaker (audio output unit), 100, 200 Echo canceller device, 101 Echo canceling unit (adaptive filter unit), 101a Adaptive filter, 101b Subtractor, 102 Double Talk determiner (DT determiner), 103 FFT processing unit, 104,201 suppression amount estimation unit, 105 ERLE calculation unit, 106 pERL calculation unit, 107 pERL decrease degree calculation unit, 108 suppression amount calculation unit, 109 residual echo suppression processing 110, IFFT processing unit, 202 noise level calculation unit.

Claims (6)

An echo canceller device used in an audio communication system having a microphone and a speaker,
An echo canceler that removes the pseudo echo component from the microphone input signal and outputs a residual signal;
A pERL calculator for calculating a pERL value indicating a ratio between the microphone input signal and the residual signal;
ERLE for calculating an ERLE value indicating a ratio of an echo signal based on an echo input from the speaker to the microphone of the microphone input signal and a residual echo signal obtained by subtracting the pseudo echo component from the echo signal. A calculation unit;
A pERL decrease degree calculation unit for calculating a decrease degree indicating a difference between the ERLE value and the pERL value;
When the value indicating the degree of decrease as a linear value is K and the value indicating the ERLE value as a linear value is T, the residual echo suppression amount is calculated from the equation (K-1) T / K (T-1). A suppression amount calculation unit for calculating
An echo canceller apparatus comprising: a residual echo suppression processing unit that generates an output signal by multiplying the residual signal by the residual echo suppression amount. An FFT processing unit for generating a frequency domain residual signal and a frequency domain microphone input signal by performing a fast Fourier transform on the residual signal and the microphone input signal output from the echo cancellation unit;
The echo canceller according to claim 1, further comprising: an IFFT processing unit that converts the output signal generated by the residual echo suppression processing unit into an output signal in a time domain. A double-talk determiner that determines whether or not a far-end voice and a near-end voice are input to the microphone ;
When it is determined not to be the double talk state by the previous SL double talk determiner, the ERLE calculation unit calculates the ERLE value, the pERL reduction degree calculation section, the decrease from the ERLE value and the pERL value Calculate degree
The echo canceller according to claim 1 or 2. A noise level calculation unit for calculating a noise level included in the microphone input signal;
4. The ERLE calculation unit calculates the ERLE value from a signal obtained by subtracting the noise level from the residual signal output from the echo canceling unit. 5. Echo canceller device. An echo cancellation method used in an audio communication system having a microphone and a speaker,
Removing a pseudo echo component from the microphone input signal and outputting a residual signal;
Calculating a pERL value indicating a ratio between the microphone input signal and the residual signal;
Calculating an ERLE value indicating a ratio of an echo signal based on an echo input from the speaker to the microphone of the microphone input signal and a residual echo signal obtained by subtracting the pseudo echo component from the echo signal; When,
Calculating a degree of decrease indicating a difference between the ERLE value and the pERL value;
When the value indicating the degree of decrease as a linear value is K and the value indicating the ERLE value as a linear value is T, the residual echo suppression amount is calculated from the equation (K-1) T / K (T-1). Calculating steps,
An echo cancellation method comprising: generating an output signal by multiplying the residual signal by the residual echo suppression amount. A process of removing a pseudo echo component from a microphone input signal in a voice communication system having a microphone and a speaker and outputting a residual signal;
Processing for calculating a pERL value indicating a ratio between the microphone input signal and the residual signal;
Processing for calculating an ERLE value indicating a ratio of an echo signal based on an echo input from the speaker to the microphone of the microphone input signal and a residual echo signal obtained by subtracting the pseudo echo component from the echo signal When,
Processing for calculating a degree of decrease indicating a difference between the ERLE value and the pERL value;
When the value indicating the degree of decrease as a linear value is K and the value indicating the ERLE value as a linear value is T, the residual echo suppression amount is calculated from the equation (K-1) T / K (T-1). A process of calculating
An echo cancellation program for causing a computer to execute a process of generating an output signal by multiplying the residual signal by the residual echo suppression amount.

Images