JP6537997B2 - Echo suppressor, method thereof, program, and recording medium - Google Patents

Description

  The present invention relates to a technique for suppressing a sound (sound echo) that has traveled from a speaker to a microphone in a hands-free communication device.

  Patent Document 1 is known as a prior art of echo suppression technology.

  FIG. 1 shows a configuration in which an echo cancellation device using an adaptive filter is provided at the front stage of the echo suppression device of Patent Document 1. The outline of the process of the configuration of FIG. 1 will be described below.

The adaptive filter unit 91 performs filtering on the reception signal x (t) (hereinafter, also simply referred to as a reception signal of the reproduction unit 2) input to the reproduction unit 2 using an adaptive filter, and estimates an echo component value y. Find the ₁ (t). Here, t is an index indicating a discrete time. The subtraction unit 92 subtracts the estimated value y ′ ₁ (t) of the echo component from the sound collection signal y ₁ (t) collected by the sound collection means 3 to obtain an error signal y ₂ (t). The adaptive filter unit 91 updates the filter coefficient of the adaptive filter using the reception signal x (t) and the error signal y ₂ (t). The frequency domain conversion unit 93 converts the time domain error signal y ₂ (t) into the frequency domain error signal Y ₂ (ω). Here, ω is an index indicating a frequency. The frequency domain conversion unit 94 converts the reception signal x (t) into the reception signal X (ω) in the frequency domain. The acoustic coupling amount estimation unit 95 estimates the acoustic coupling amount D (ω) using the error signal Y ₂ (ω) in the frequency domain and the reception signal X (ω) in the frequency domain. The echo level estimation unit 96 multiplies the level of the reception signal X (ω) in the frequency domain by the amount of acoustic coupling to estimate the level of the echo component included in the error signal Y ₂ (ω) in the frequency domain. The gain acquisition unit 97 acquires the gain G (ω) using the estimated value R (ω) of the level of the echo component and the error signal Y ₂ (ω) in the frequency domain. The multiplication unit 98 multiplies the error signal Y ₂ (ω) in the frequency domain by the gain G (ω) to obtain the transmission signal Y ₃ (ω) in the frequency domain. The time domain conversion unit 99 converts the transmission signal Y ₃ (ω) in the frequency domain into the transmission signal y ₃ (t) in the time domain, and outputs it as an output value of the echo canceller 90.

  With such a configuration, in the configuration of Patent Document 1, the residual echo that can not be erased by the echo cancellation apparatus using the adaptive filter at the previous stage is suppressed.

JP 2008-5094 A

  If the echo cancellation of the echo cancellation apparatus using the preceding adaptive filter is stable, the echo suppression apparatus of Patent Document 1 can suppress the residual echo.

  However, when an excessive reception signal that causes distortion in the reproduced sound by the speaker unit or the speaker amplifier of the reproduction means 2 is input to the echo suppression apparatus of Patent Document 1, the distortion component is an echo using the adaptive filter of the previous stage. The erasing device can not erase the signal, resulting in a large residual echo. The conventional echo suppressor can not cope with the rapidly increasing residual echo, resulting in a problem that the suppression is not sufficiently performed. Even when the echo suppression apparatus of Patent Document 1 is used alone as an echo suppression apparatus without providing an echo cancellation apparatus using an adaptive filter in the previous stage, it can not cope with the rapidly increasing echo as well. There is a problem that is not done enough.

  The present invention provides an echo suppression apparatus capable of stably suppressing an echo even when distortion occurs in a speaker unit or a speaker amplifier by inputting an excessive reception signal, a method, a program, and a recording medium. is there.

  In order to solve the above-mentioned problems, according to one aspect of the present invention, an echo suppression apparatus is provided with a reproduction means and a sound collection means from the ratio of a value based on a collected sound signal in the frequency domain to a reception signal in the frequency domain. And the level of the reception signal in the frequency domain multiplied by the amount of acoustic coupling to calculate the level of the echo component contained in the collected signal for each frequency band. When reproducing the reception signal in the reproduction means using the echo level estimation unit to be estimated, the level of the reception signal, the estimated value of the level of the echo component, and the level of the collected signal, the level of the reception signal is large If there is a possibility that distortion occurs in the reproduced sound, a gain acquisition unit for obtaining, for each frequency, a gain G (ω) having a larger amount of suppression than the gain used if distortion does not temporarily occur, and sound collection in the frequency domain Gay to value based on signal And an echo suppression unit multiplied by the G (ω).

  In order to solve the above problems, according to another aspect of the present invention, an echo suppression method comprises: reproducing means and sound collection from the ratio of a value based on a collected sound signal in the frequency domain to a reception signal in the frequency domain An acoustic coupling amount estimation step of estimating the acoustic coupling amount between the means and the frequency domain, and multiplying the level of the reception signal in the frequency domain by the acoustic coupling amount to obtain the level of the echo component included in the collected signal per frequency band When the receiving signal is reproduced by the reproduction means using the echo level estimation step to be estimated, the level of the receiving signal, the estimated value of the level of the echo component, and the level of the collected signal, the level of the receiving signal is If distortion is likely to occur in the reproduced sound because it is large, a gain acquisition step for obtaining for each frequency a gain G (ω) having a larger amount of suppression than a gain used if distortion does not temporarily occur, and sound Comprising a value based on the No. and echo suppression step of multiplying a gain G (ω).

  According to the present invention, it is possible to stably suppress an echo even if an excessive reception signal is input and distortion occurs in the speaker unit or the speaker amplifier.

FIG. 1 is a functional block diagram of a prior art echo canceler. FIG. 2 is a functional block diagram of an echo cancellation apparatus according to the first embodiment. The figure which shows the example of the processing flow of the echo cancellation apparatus which concerns on 1st embodiment. FIG. 2 is a functional block diagram of a gain acquisition unit according to the first embodiment. The figure which shows the example of the processing flow of the gain acquisition part which concerns on 1st embodiment. FIG. 8 is a functional block diagram of a gain acquisition unit according to a second embodiment. The figure which shows the example of the processing flow of the gain acquisition part which concerns on 2nd embodiment. The functional block diagram of the gain acquisition part which concerns on 3rd embodiment. The figure which shows the example of the processing flow of the gain acquisition part which concerns on 3rd embodiment.

  Hereinafter, embodiments of the present invention will be described. In the drawings used in the following description, the same reference numerals are given to constituent parts having the same functions and steps for performing the same processing, and redundant description will be omitted. The processing performed for each element of a vector or matrix is applied to all elements of the vector or matrix unless otherwise noted.

First Embodiment
FIG. 2 shows a functional block diagram of the echo suppressor 100 according to the first embodiment, and FIG. 3 shows an example of its processing flow.

  The echo suppression apparatus 100 includes an adaptive filter unit 101, a subtraction unit 102, a frequency domain conversion unit 103, a frequency domain conversion unit 104, an acoustic coupling amount estimation unit 105, an echo level estimation unit 106, an excessive level detection unit 110, and a gain acquisition unit 120. , And an echo suppression unit 108 and a time domain conversion unit 109.

The echo suppressor 100 receives the reception signal x (t) reproduced by the reproduction means 2 and the collected signal y ₁ (t) collected by the sound collecting means 3 as an input, and the echo signal y ₁ (t) A transmission signal y ₃ (t) obtained by eliminating and suppressing the estimated value of the component is obtained and output. The processing content of each part will be described below.

The reproduction means 2 is composed of a speaker, a speaker unit, a speaker amplifier, etc., and reproduces the received signal x (t). The sound collection means 3 is composed of a microphone or the like, and outputs a sound collection signal y ₁ (t).

<Adaptive filter unit 101>
The adaptive filter unit 101 receives the reception signal x (t) of the reproduction means 2 and the error signal y ₂ (t), and uses these values to be included in the sound collection signal y ₁ (t) of the sound collection means 3. An estimated value y ′ ₁ (t) of the echo component to be obtained is obtained (S 101) and output.

For example, using the reception signal x (t) and a filter coefficient H (t) described later, an estimated value y ′ ₁ (t) is obtained by the following equation.
y ' ₁ (t) = H (t) ^T X (t) (1)
H (t) = (h (0), h (1), ..., h (L-1)) ^T (2)
X (t) = (x (t), x (t-1), ..., x (t-L + 1)) ^T (3)
However, the superscript T denotes the transpose, A ^T denotes the transpose of a vector A, L represents a tap length of the adaptive filter.

Here, the filter coefficient H (t) is updated in a filter coefficient update unit (not shown) inside the adaptive filter unit 101. For example, when using the NLMS algorithm, the filter coefficient H (t) is updated by the following equation.
H (t + 1) = H (t) + aX (t) y 2 (t) / X (t) T X (t) (4)
0 <a <2 (5)
Where a represents the step size of the NLMS algorithm. The method of updating or determining the filter coefficient H (t) is not limited to this method, and a conventional method may be used.

<Subtractor 102>
The subtraction unit 102 receives the sound collection signal y ₁ (t) of the sound collection means 3 and the estimated value y ′ ₁ (t) of the echo component, and obtains the difference y ₁ (t) −y ′ ₁ (t) (S102) The error signal y ₂ (t) (= y ₁ (t) −y ′ ₁ (t)) is output.

<Frequency domain transform unit 103 and frequency domain transform unit 104>
The frequency domain conversion unit 103 receives the error signal y ₂ (t), converts it into an error signal Y ₂ (ω) in the frequency domain (S 103), and outputs it. As a conversion method, FFT (short time Fourier transform) or the like can be used.

  The frequency domain conversion unit 104 receives the reception signal x (t), converts it to the reception signal X (ω) in the frequency domain using the same conversion method as the frequency domain conversion unit 103 (S104), and outputs it.

<Acoustic Coupling Amount Estimation Unit 105>
The acoustic coupling amount estimation unit 105 receives the error signal Y ₂ (ω) in the frequency domain and the reception signal X (ω) in the frequency domain, and determines the ratio between the error signal Y ₂ (ω) and the reception signal in the frequency domain. The acoustic coupling amount D (ω) between the reproduction means and the sound collection means is estimated for each frequency domain (S105) and output. For example, the acoustic coupling amount D (ω) is an amplitude value of the transfer characteristic between the reproducing means 2 and the sound collecting means 3, and the error signal Y ₂ (ω) in the frequency domain and the received signal X (ω) in the frequency domain The ratio of the absolute value of In addition, time smoothing is performed to improve the accuracy of the amount of acoustic coupling. The acoustic coupling amount D (ω) is obtained by the following equation.
D (ω) = E {| Y ₂ (ω) | / | X (ω) |} (6)
However, E {A} represents taking an average value of A, and | A | represents taking an absolute value of A.

<Echo level estimation unit 106>
The echo level estimation unit 106 receives the reception signal X (ω) in the frequency domain and the acoustic coupling amount D (ω), and multiplies the level of the reception signal X (ω) in the frequency domain by the acoustic coupling amount D (ω). The level of the echo component included in the collected signal is estimated for each frequency band (S106), and an estimated value R (ω) is output.

For example, when the room echo is ignored, the level of the echo component can be estimated by multiplying the reception signal X (ω) by the acoustic coupling amount D (ω). However, since there is room sound in practice, it is necessary to estimate the echo component including the echo component. Usually, since the acoustic component of the room decays exponentially with time, the level of the echo component is estimated by the following equation.
R (ω) = D (ω) · P (ω)
P (ω) = | X (ω) | for P '(ω) ≦ | X (ω) |
P (ω) = u · P ′ (ω) + (lu) · | X (ω) | for P ′ (ω)> | X (ω) | (7)
Where P (ω) is the received signal after time smoothing corresponding to the echo, P ′ (ω) is P (ω) one frame before, and u is the echo length (reverberation time) It is a coefficient for adjusting the assumed value of and a fixed value is set in advance. For example, u takes a value of 0 ≦ u <1, and an environment with longer reverberation time is simulated as closer to 1 and an environment with shorter reverberation time is simulated as closer to 0.

<Excessive Level Detection Unit 110>
The excessive level detection unit 110 receives the reception signal x (t) in the time domain, obtains the level s (t) of the reception signal x (t) (S110), and outputs it. The level s (t) of the reception signal x (t) is the absolute value s (t) = | x (t) | of the reception signal x (t) or the signal s (t) =. Alpha. The maximum value holding level of the reception signal x (t) calculated using the following equation is used: s ′ (t) + (l−α) · x (t) |
s (t) = | x (t) | for s' (t) ≦ | x (t) |
s (t) = α · s ′ (t) + (l−α) · x (t) | for s ′ (t)> | x (t) |
s (t) is a received signal after time smoothing corresponding to the echo, s' (t) is s (t) one frame before, α is a smoothing function, 0 to 1 Take the value between

<Gain Acquisition Unit 120>
Gain acquisition section 120 receives estimated value R (ω) of the level of the echo component, level s (t) of reception signal x (t), and error signal Y ₂ (ω) in the frequency domain, and receives reception signal x If there is a possibility that distortion occurs in the reproduced sound because the level s (t) of (t) is large, gain G (ω) with a larger amount of suppression for each frequency than the gain used if distortion does not occur temporarily Determine (S120) and output.

  FIG. 4 shows a functional block diagram of the gain acquisition unit 120, and FIG. 5 shows an example of its processing flow.

The gain acquisition unit 120 includes a normal time multiplication coefficient storage unit 121, an excessive time multiplication coefficient storage unit 122, a coefficient selection unit 123, a coefficient multiplication unit 124, and an echo suppression gain acquisition unit 125.
(Normal multiplication coefficient storage unit 121 and excessive multiplication coefficient storage unit 122)
The normal multiplication coefficient storage unit 121 and an excessive time multiplier coefficient storage unit 122, previously stored normal multiplication coefficient gamma ₁ and excessive time multiplication factor gamma _2, respectively. It is assumed that γ ₁ <γ ₂ .

(Coefficient selection unit 123)
The coefficient selection unit 123 receives the level s (t) of the reception signal x (t), and if the level s (t) of the reception signal x (t) is large, distortion may occur in the reproduced sound. taking out an excessive time multiplication coefficient gamma ₂ from the time of multiplying the coefficient storage unit 122, and outputs the coefficient multiplication unit 124 (S123). Also, if the strain does not occur, it is taken out normal multiplication coefficient gamma ₁ from normal multiplication coefficient storage unit 121, and outputs the coefficient multiplication unit 124 (S123). For example, if the level of the received signal x (t) s (t) exceeds the threshold value beta ₁ set in advance, there is a possibility that distortion occurs in the reproduced sound (hereinafter, also referred to the level excessive level) determination Do. The threshold value β ₁ may be checked in advance by experiment, simulation or the like in accordance with the regeneration means 2.

(Coefficient multiplication unit 124)
Coefficient multiplying unit 124, the estimated value of the level of the echo component R (omega), the product by multiplying the excess time multiplication factor gamma ₂ or normal multiplication coefficient gamma ₁ selected in the coefficient selection unit 123 (γ ₁ R ( ω) or γ ₂ R (ω)) is obtained (S 124) and output. The multiplied signal (γ ₁ R (ω) or γ ₂ R (ω)) is used as an estimate of the level of the echo component. Since γ ₁ <γ ₂ indicates that distortion may occur, the level of the echo component will be highly estimated.

(Echo suppression gain acquisition unit 125)
The echo suppression gain acquisition unit 125 receives the product (γ ₁ R (ω) or γ ₂ R (ω)) and the error signal Y ₂ (ω) in the frequency domain, and calculates the product (γ ₁ R (R) for each frequency band. comparing the level of omega) or γ ₂ R (ω)) and the error signal Y ₂ (omega), and set a large gain suppression amount as the product is large, the gain G (omega used in the echo suppressing unit 108 to be described later ) And output (S125).

For example, the gain can be set by the same method as that of Patent Document 1. First, when the error signal Y ₂ (ω) contains many echo components, the level of the product (γ ₁ R (ω) or γ ₂ R (ω)) and the level of the error signal Y ₂ (ω) are close Therefore, the level of the error signal Y ₂ (ω) is smaller than the product (γ ₁ R (ω) or γ ₂ R (ω)) multiplied by a preset fixed value, for example, a fixed coefficient C of 1 or more. In this case, it is detected as a period in which a large amount of echo component is contained. For example, assuming that the level of the error signal Y ₂ (ω) is W (ω), this condition is expressed by the following equation.
W (ω) ≦ C · γR (ω) (8)
However, γ is γ ₁ or γ ₂ . As the level W (omega) of the error signal Y ₂ (omega), the absolute value and the error signal Y ₂ (omega), the absolute value may be used smoothed signal. For example,
W (ω) = | Y (ω) | for W '(ω) ≦ | Y (ω) |
W (ω) = u · W ′ (ω) + (lu) · | Y (ω) | for W ′ (ω)> | Y (ω) |
I assume. Where W (ω) is the received signal after time smoothing corresponding to echo, W '(ω) is W (ω) one frame before, and u is the echo length (reverberation time) It is a coefficient for adjusting the assumed value of and a fixed value is set in advance. For example, u takes a value of 0 ≦ u <1, and an environment with longer reverberation time is simulated as closer to 1 and an environment with shorter reverberation time is simulated as closer to 0.

When it is detected as a period in which a large amount of echo components are contained, the instantaneous gain coefficient g (ω) of that band is set to an echo suppression amount D set in advance as a fixed value. However, the amount of echo suppression D takes, for example, a value of 0 ≦ D <1, and the smaller the value, the more the echo suppression amount increases, but the deterioration of the near-end speaker voice at the time of double talk increases. Next, when the echo component is not detected as a period including a large amount of echo components, the echo component is small, so the instantaneous gain coefficient g (ω) is set to a preset fixed value, for example, 1 and the error signal Y ₂ (ω Let pass). This gain control can be expressed by the following equation.
g (ω) = D for W (ω) ≦ C · R (ω)
g (ω) = 1 for W (ω)> C · R (ω)
Next, the instantaneous gain coefficient g (ω) is time-smoothed to obtain a gain G (ω) to be output to the echo suppression unit 108. By smoothing the time, it is possible to suppress the sound quality deterioration due to the abrupt change of the gain. Temporal smoothing is performed as follows, for example.
G (ω) = a · G ′ (ω) + (la) · g (ω) for g (ω) ≦ G ′ (ω)
G (ω) = b · G ′ (ω) + (lb) · g (ω) for g (ω)> G ′ (ω) (9)
Here, G ′ (ω) is a gain G (ω) one frame before. a is a smoothing coefficient at the time of gain decrease, b is a smoothing coefficient at the time of gain increase, and is set in advance as a fixed value. a and b take values between 0 and 1, and the closer to 1, the longer time is smoothed, and the closer to 0, the shorter time is smoothed.

The above gain acquisition method is an example, and the product (γ ₁ R (ω) or γ ₂ R (ω)) is compared with the level of the error signal Y ₂ (ω) for each frequency band, Other methods may be used as long as the larger the product, the larger the amount of suppression can be set. For example, g (ω) may be used as gain G (ω) as it is without performing time smoothing.

<Echo Suppression Unit 108>
Echo suppressing unit 108 receives the error signal Y ₂ in the frequency domain and (omega) and a gain G (omega), multiplied by the gain G (omega) the error signal Y ₂ (omega) of the frequency domain, transmission signal Y ₃ (ω) (Y ₃ (ω) = G (ω) Y ₂ (ω)) is obtained (S 108) and output.

<Time domain conversion unit 109>
The time domain conversion unit 109 receives the transmission signal Y ₃ (ω), converts it into a transmission signal y ₃ (t) in the time domain (S 109), and outputs it. As a conversion method, a method corresponding to the conversion method used in the frequency domain conversion unit 103 and the frequency domain conversion unit 104 may be used. For example, IFFT (inverse short time Fourier transform) or the like can be used.

<Effect>
With the above configuration, a value close to the echo level including an increase in echo due to distortion of the speaker unit or the speaker amplifier is estimated by multiplying the echo level by a large coefficient only when an excessive reception signal is detected. It is possible to perform stable echo suppression.

<Modification>
The adaptive filter unit 101 and the subtraction unit 102 may not necessarily be included. In that case, in the processing after S103, the collected sound signal y ₁ (t) may be used instead of the error signal y ₂ (t). Incidentally, it can be said from the error signal y ₂ (t) are collected signal y ₁ (t) is a value obtained by subtracting the estimated value y _'1 (t) of the echo component, a value based on the collected signal y ₁ (t) . Of course, the collected signal y ₁ (t) itself can also be said to be a value based on the collected signal y ₁ (t).

In the present embodiment, the excessive level detection unit 110 only determines the level s (t) of the reception signal x (t), but determines whether the level s (t) exceeds the threshold β ₁ and determines the determination result. It may be configured to output. In coefficient selection section 123, whether the determination result (level s (t) exceeds threshold β ₁ or not, in other words, distortion may occur in the reproduced sound because the level s (t) of reception signal x (t) is large. The coefficient may be selected according to the judgment result indicating whether or not there is a sex.

Second Embodiment
Description will be made focusing on parts different from the first embodiment.
In the second embodiment, a gain acquisition unit 220 is included instead of the gain acquisition unit 120.

<Gain Acquisition Unit 220>
Gain acquisition section 220 receives estimated value R (ω) of the level of the echo component, level s (t) of reception signal x (t), and error signal Y ₂ (ω) in the frequency domain, and receives reception signal x If there is a possibility that distortion occurs in the reproduced sound because the level s (t) of (t) is large, gain G (ω) with a larger amount of suppression for each frequency than the gain used if distortion does not occur temporarily Determine (S220) and output.

  FIG. 6 shows a functional block diagram of the gain acquisition unit 220, and FIG. 7 shows an example of its processing flow.

  The gain acquisition unit 220 includes an echo suppression gain acquisition unit 225, an excessive gain storage unit 226, and a gain selection unit 227.

(Echo suppression gain acquisition unit 225)
The echo suppression gain acquisition unit 225 receives the estimated value R (ω) of the level of the echo component and the error signal Y ₂ (ω), and the estimated value R (ω) of the level of the echo component and the error signal Y ₂ (ω) The gain G (1, .omega.) Having a larger suppression amount is set for each frequency band as the estimated value R (.omega.) Of the level of the echo component is larger for each frequency band (S225). The specific process is the same as that of the echo suppression gain acquisition unit 125, and uses the estimated value R (ω) of the level of the echo component instead of the product (γ ₁ R (ω) or γ ₂ R (ω)). The point is different.

(Excessive gain storage unit 226)
The excessive gain G storage unit 226 stores the excessive gain G (2, ω) in advance. Note that the over-time gain G (2, ω) is satisfied so that G (2, ω) <G (1, ω) is satisfied for any G (1, ω) obtained by the echo suppression gain acquisition unit 225. Set). The point is that the amount of suppression G is larger than the gain G (1, ω).

(Gain selection unit 227)
The gain selection unit 227 receives the level s (t) of the reception signal x (t) obtained by the excessive level detection unit 110, and the level of the level s (t) of the reception signal x (t) is large. If distortion is likely to occur, the in-excess gain G (2, ω) is taken out from the in-excess gain storage unit 226 and is output as the gain G (ω) used by the echo suppression unit 108 (S227). In this case, the gain G (1, ω) obtained by the echo suppression gain acquisition unit 225 is output as the gain G (ω) used by the echo suppression unit 108 (S227). For example, it is determined that if the level of the received signal x (t) s (t) exceeds the threshold value beta ₂ set in advance, there is a possibility that distortion occurs in the reproduced sound. Threshold beta ₂ experiments in accordance with the reproduction unit 2, it is sufficient to examine in advance by simulation or the like.

<Effect>
With such a configuration, the same effect as that of the first embodiment can be obtained. A modification of the first embodiment may be combined with the present embodiment.

For example, in the present embodiment, the excessive level detector 110, but just finding the level s (t) of the received signal x (t), determines whether the level s (t) exceeds the threshold value beta _2, determination The result may be output. In gain selection section 227, whether or not the determination result (level s (t) exceeds threshold β ₂ or not, in other words, distortion may occur in the reproduced sound because level s (t) of reception signal x (t) is large. The gain may be selected according to the judgment result indicating whether or not there is a sex.

Third Embodiment
Description will be made focusing on parts different from the first embodiment. The third embodiment is a configuration in which the first embodiment and the second embodiment are combined.
In the third embodiment, a gain acquisition unit 320 is included instead of the gain acquisition unit 120.

<Gain Acquisition Unit 320>
Gain acquisition section 320 receives estimated value R (ω) of the level of the echo component, level s (t) of received signal x (t), and error signal Y ₂ (ω) in the frequency domain, and receives received signal x If there is a possibility that distortion occurs in the reproduced sound because the level s (t) of (t) is large, gain G (ω) with a larger amount of suppression for each frequency than the gain used if distortion does not occur temporarily Obtain (S320) and output.

  FIG. 8 shows a functional block diagram of the gain acquisition unit 320, and FIG. 9 shows an example of its processing flow.

  The gain acquisition unit 320 is a normal multiplication coefficient storage unit 121, an excess multiplication coefficient storage unit 122, a coefficient selection unit 123, a coefficient multiplication unit 124, an echo suppression gain acquisition unit 325, an excess gain storage unit 226, and a gain selection unit 227. including.

Note that the threshold value beta ₂ used in the threshold beta ₁ and gain selection unit 227 used in the coefficient selector 123 is set to be a β ₁ <β _2.

(Echo suppression gain acquisition unit 325)
The echo suppression gain acquisition unit 325 receives the product (γ ₁ R (ω) or γ ₂ R (ω)) and the error signal Y ₂ (ω) in the frequency domain, and calculates the product (γ ₁ R (R) for each frequency band. comparing the level of omega) or γ ₂ R (ω)) and the error signal Y ₂ (omega), and set a large gain suppression amount as the product is large, the gain G (1, omega) the gain selector 227 (S325).

The processing contents of the excessive gain storage unit 226 and the gain selection unit 227 are the same as in the second embodiment. At this time, in the case of β ₂ <s (t), it is clear that the gain G (2, ω) is selected in the gain selection unit 227, so the coefficient selection unit 123, the coefficient multiplication unit 124, the echo The processing of the suppression gain acquisition unit 325 may be omitted.

<Effect>
With such a configuration, the same effects as those of the first embodiment and the second embodiment can be obtained. In the present embodiment, in the case of less than the threshold value beta ₂ exceeds the threshold value beta ₁ multiplied by the excess time multiplication coefficient to the echo component, as shown in the first embodiment, if the threshold is exceeded beta ₂ as shown in the second embodiment Force the gain to be replaced with the overtime gain.

  By doing this, when the distortion is not so large with an excessive reception signal, the estimated echo is largely estimated, and when the distortion is large, the transmission voice is completely suppressed. A modification of the first embodiment may be combined with the present embodiment.

For example, in the present embodiment, the excessive level detector 110, but just finding the level s (t) of the received signal x (t), determines whether the level s (t) exceeds the threshold value beta ₁ or beta ₂ And the determination result may be output. In coefficient selection section 123, whether the determination result (level s (t) exceeds threshold β ₁ or not, in other words, distortion may occur in the reproduced sound because the level s (t) of reception signal x (t) is large. The coefficient may be selected according to the judgment result indicating whether or not there is a sex. In gain selection section 227, whether or not the determination result (level s (t) exceeds threshold β ₂ or not, in other words, distortion may occur in the reproduced sound because level s (t) of reception signal x (t) is large. The gain may be selected according to the judgment result indicating whether or not there is a sex.

<Other Modifications>
The present invention is not limited to the above embodiments and modifications. For example, the various processes described above may be performed not only in chronological order according to the description, but also in parallel or individually depending on the processing capability of the apparatus that executes the process or the necessity. In addition, changes can be made as appropriate without departing from the spirit of the present invention.

<Program and Recording Medium>
In addition, various processing functions in each device described in the above-described embodiment and modification may be realized by a computer. In that case, the processing content of the function that each device should have is described by a program. By executing this program on a computer, various processing functions in each of the above-described devices are realized on the computer.

  The program describing the processing content can be recorded in a computer readable recording medium. As the computer readable recording medium, any medium such as a magnetic recording device, an optical disc, a magneto-optical recording medium, a semiconductor memory, etc. may be used.

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

  For example, a computer that executes such a program first temporarily stores a program recorded on a portable recording medium or a program transferred from a server computer in its own storage unit. Then, at the time of execution of the process, the computer reads the program stored in its storage unit and executes the process according to the read program. In another embodiment of the program, the computer may read the program directly from the portable recording medium and execute processing in accordance with the program. Furthermore, each time a program is transferred from this server computer to this computer, processing according to the received program may be executed sequentially. In addition, a configuration in which the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes processing functions only by executing instructions and acquiring results from the server computer without transferring the program to the computer It may be Note that the program includes information provided for processing by a computer that conforms to the program (such as data that is not a direct command to the computer but has a property that defines the processing of the computer).

  In addition, although each device is configured by executing a predetermined program on a computer, at least a part of the processing content may be realized as hardware.

Claims (8)

An acoustic coupling amount estimation unit for estimating an acoustic coupling amount between the reproduction means and the sound pickup means for each frequency domain from a ratio between a value based on a collected sound signal in the frequency domain and a reception signal in the frequency domain;
An echo level estimation unit that estimates the level of the echo component included in the collected signal by multiplying the level of the reception signal in the frequency domain by the acoustic coupling amount;
When the reproduction signal is reproduced by the reproduction means using the level of the reception signal, the estimated value of the level of the echo component, and the level of the collected signal, the level of the reception signal is large. A gain acquisition unit which obtains, for each frequency, a gain G (ω) having a larger amount of suppression than a gain used if distortion does not occur if distortion may occur in the reproduced sound;
An echo suppression unit that multiplies the gain G (ω) by a value based on the collected sound signal in the frequency domain;
The gain acquisition unit
If the normal multiplication coefficient is smaller than the excessive multiplication coefficient, and if there is a possibility that distortion occurs in the reproduced sound because the level of the reception signal is large, the excessive multiplication coefficient is selected, and no distortion occurs. And a coefficient selection unit for selecting a multiplication coefficient at a normal time,
A coefficient multiplication unit which obtains a product by multiplying the estimated value of the level of the echo component and the over-time multiplication coefficient or the normal-time multiplication coefficient selected in the coefficient selection unit;
An echo suppression gain acquisition unit which compares, for each frequency band, the product with the level of a value based on the collected sound signal, sets a larger gain as the product is larger, and sets the gain as the gain G (ω); including,
Echo suppressor. An acoustic coupling amount estimation unit for estimating an acoustic coupling amount between the reproduction means and the sound pickup means for each frequency domain from a ratio between a value based on a collected sound signal in the frequency domain and a reception signal in the frequency domain;
An echo level estimation unit that estimates the level of the echo component included in the collected signal by multiplying the level of the reception signal in the frequency domain by the acoustic coupling amount;
When the reproduction signal is reproduced by the reproduction means using the level of the reception signal, the estimated value of the level of the echo component, and the level of the collected signal, the level of the reception signal is large. A gain acquisition unit which obtains, for each frequency, a gain G (ω) having a larger amount of suppression than a gain used if distortion does not occur if distortion may occur in the reproduced sound;
An echo suppression unit that multiplies the gain G (ω) by a value based on the collected sound signal in the frequency domain;
The gain acquisition unit
The estimated value of the level of the echo component is compared with the level of the value based on the collected sound signal, and the larger the estimated value of the level of the echo component, the larger the gain G (1, ω) of the suppression amount is for each frequency band An echo suppression gain acquisition unit to be set;
It is assumed that the excessive gain G (2, ω) has a larger suppression amount than the gain G (1, ω), and if the level of the reception signal is large, distortion may occur in the reproduced sound. Selecting a time gain G (2, ω), and selecting no gain G (1, ω) if distortion does not occur;
Echo suppressor. An acoustic coupling amount estimation unit for estimating an acoustic coupling amount between the reproduction means and the sound pickup means for each frequency domain from a ratio between a value based on a collected sound signal in the frequency domain and a reception signal in the frequency domain;
An echo level estimation unit that estimates the level of the echo component included in the collected signal by multiplying the level of the reception signal in the frequency domain by the acoustic coupling amount;
When the reproduction signal is reproduced by the reproduction means using the level of the reception signal, the estimated value of the level of the echo component, and the level of the collected signal, the level of the reception signal is large. A gain acquisition unit which obtains, for each frequency, a gain G (ω) having a larger amount of suppression than a gain used if distortion does not occur if distortion may occur in the reproduced sound;
An echo suppression unit that multiplies the gain G (ω) by a value based on the collected sound signal in the frequency domain;
The gain acquisition unit
It is assumed that β ₁ <β ₂ , the level of the reception signal is s (t), the normal time multiplication factor is smaller than the overtime multiplication factor, and if β ₁ <s (t) ≦ β ₂ , the overtime multiplication is performed A coefficient selection unit which selects a coefficient, and in the case of s (t) ≦ β ₁ selects a normal multiplication coefficient;
A coefficient multiplication unit which obtains a product by multiplying the estimated value of the level of the echo component and the over-time multiplication coefficient or the normal-time multiplication coefficient selected in the coefficient selection unit;
An echo suppression gain acquisition unit that compares, for each frequency band, the product with the level of a value based on the collected signal, and sets a gain G (1, ω) with a larger amount of suppression as the product is larger;
The overrun gain G (2, ω) is larger than the gain G (1, ω) by the amount of suppression, and in the case of β ₂ <s (t), the overrun gain G (2, ω) is selected, in the case of s (t) ≦ β ₂ selects the gain G (1, ω), and a gain selection unit for the gain G (ω),
Echo suppressor. An acoustic coupling amount estimation step of estimating, for each frequency domain, an acoustic coupling amount between the reproduction means and the sound collecting means from a ratio between a value based on a collected sound signal in the frequency domain and a reception signal in the frequency domain;
An echo level estimation step of estimating the level of the echo component included in the collected signal by multiplying the level of the reception signal in the frequency domain by the acoustic coupling amount;
When the reproduction signal is reproduced by the reproduction means using the level of the reception signal, the estimated value of the level of the echo component, and the level of the collected signal, the level of the reception signal is large. A gain acquisition step of obtaining, for each frequency, a gain G (ω) having a larger amount of suppression than a gain used if distortion does not occur if distortion may occur in the reproduced sound;
Look including the echo suppression step of multiplying the gain G (omega) the value based on the collected sound signal in the frequency domain,
The gain acquisition step is
If the normal multiplication coefficient is smaller than the excessive multiplication coefficient, and if there is a possibility that distortion occurs in the reproduced sound because the level of the reception signal is large, the excessive multiplication coefficient is selected, and no distortion occurs. And a coefficient selection step of selecting a multiplication coefficient normally.
A coefficient multiplication step for obtaining a product by multiplying the estimated value of the level of the echo component and the over-time multiplication coefficient or the normal-time multiplication coefficient selected in the coefficient selection step;
An echo suppression gain acquisition step of comparing the product with the level of a value based on the collected sound signal for each frequency band, setting a larger gain as the product is larger, and setting the gain as the gain G (ω); including,
Echo suppression method. An acoustic coupling amount estimation step of estimating, for each frequency domain, an acoustic coupling amount between the reproduction means and the sound collecting means from a ratio between a value based on a collected sound signal in the frequency domain and a reception signal in the frequency domain;
An echo level estimation step of estimating the level of the echo component included in the collected signal by multiplying the level of the reception signal in the frequency domain by the acoustic coupling amount;
When the reproduction signal is reproduced by the reproduction means using the level of the reception signal, the estimated value of the level of the echo component, and the level of the collected signal, the level of the reception signal is large. A gain acquisition step of obtaining, for each frequency, a gain G (ω) having a larger amount of suppression than a gain used if distortion does not occur if distortion may occur in the reproduced sound;
Look including the echo suppression step of multiplying the gain G (omega) the value based on the collected sound signal in the frequency domain,
The gain acquisition step is
The estimated value of the level of the echo component is compared with the level of the value based on the collected sound signal, and the larger the estimated value of the level of the echo component, the larger the gain G (1, ω) of the suppression amount is for each frequency band An echo suppression gain acquisition step to be set;
It is assumed that the excessive gain G (2, ω) has a larger suppression amount than the gain G (1, ω), and if the level of the reception signal is large, distortion may occur in the reproduced sound. Selecting a gain G (2, ω) and selecting a gain G (1, ω) if no distortion occurs, and selecting the gain G (ω).
Echo suppression method. An acoustic coupling amount estimation step of estimating, for each frequency domain, an acoustic coupling amount between the reproduction means and the sound collecting means from a ratio between a value based on a collected sound signal in the frequency domain and a reception signal in the frequency domain;
An echo level estimation step of estimating the level of the echo component included in the collected signal by multiplying the level of the reception signal in the frequency domain by the acoustic coupling amount;
When the reproduction signal is reproduced by the reproduction means using the level of the reception signal, the estimated value of the level of the echo component, and the level of the collected signal, the level of the reception signal is large. A gain acquisition step of obtaining, for each frequency, a gain G (ω) having a larger amount of suppression than a gain used if distortion does not occur if distortion may occur in the reproduced sound;
Look including the echo suppression step of multiplying the gain G (omega) the value based on the collected sound signal in the frequency domain,
The gain acquisition step is
It is assumed that β ₁ <β ₂ , the level of the reception signal is s (t), the normal time multiplication factor is smaller than the overtime multiplication factor, and if β ₁ <s (t) ≦ β ₂ , the overtime multiplication is performed select the coefficients, a coefficient selection step of selecting a normal multiplication factor in the case of s (t) ≦ β _1,
A coefficient multiplication step for obtaining a product by multiplying the estimated value of the level of the echo component and the over-time multiplication coefficient or the normal-time multiplication coefficient selected in the coefficient selection step;
An echo suppression gain acquisition step of comparing the product with the level of a value based on the collected sound signal for each frequency band, and setting a gain G (1, ω) with a larger amount of suppression as the product is larger;
The overrun gain G (2, ω) is larger than the gain G (1, ω) by the amount of suppression, and in the case of β ₂ <s (t), the overrun gain G (2, ω) is selected, in the case of s (t) ≦ β ₂ selects the gain G (1, ω), and a gain selection step of the gain G (ω),
Echo suppression method. A program for causing a computer to function as the echo suppressor according to any one of claims 1 to 3 . A computer-readable recording medium having recorded thereon a program for causing a computer to function as the echo suppressor according to any one of claims 1 to 3 .

Images