JP2010028653A - Echo canceling apparatus, echo canceling method, its program, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve a desired echo canceling volume with a low calculation cost even if sound directivity is changed by an adaptive beam forming processing. <P>SOLUTION: The echo canceling apparatus for deleting echo signals included in sound pickup signals picked up with N of sound pickup means, suppressing noise signals included in the sound pickup signals to output the sound pickup signals as output signals deletes elements of echo signals contained in frequency ranges lower than a predetermined reference value, outputs residual echo signals, suppresses the noise signals from N of the residual echo signals to output adaptive beam forming signals, and deletes the echo signals from the adaptive beam forming signals to output the output signals. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an echo canceling apparatus and an echo canceling method for canceling an echo signal included in a sound pickup signal picked up by a plurality of sound pickup means, suppressing a noise signal included in the sound pickup signal and outputting it as an output signal , The program, and a recording medium.

  In addition to conventional telephone conferences and video conferences, in recent years, opportunities for using hands-free loudspeaking such as video phone desktop conferences using PCs (personal computers) are increasing. In the hands-free loudspeaking call, the received voice of the other party is emitted from the reproduction means (for example, a speaker), while the voice of the receiver is picked up by the sound collection means (for example, a microphone). There is a great advantage that the minutes can be written and the PC can be operated.

  However, the voice of the other party emitted from the speaker not only reaches the receiver, but also enters the microphone and is collected, which deteriorates the call quality as an acoustic echo and also causes howling. In order to eliminate such echo signal components and prevent the occurrence of howling, an echo cancellation technique has been introduced in hands-free loudspeaking calls.

  In addition, there is adaptive beamforming technology that picks up a speaker's voice with high accuracy after selectively receiving and following the user's voice from various environmental noises and performing noise suppression. This technique is described in the sound collection method and sound collection device “Echo Suppression by Microphone Array, AGC Microphone Array for Each Direction” of Patent Document 1. This technique suppresses a noise signal from a noise source by controlling acoustic directivity using a plurality of microphones, and collects a speaker's voice with high accuracy. In recent years, this adaptive beamforming technique is also provided in an echo canceller (see Non-Patent Document 1).

FIG. 1 shows an example of the functional configuration of a conventional echo canceling apparatus 100 equipped with adaptive beamforming. The echo canceller 100 is connected to N (N is an integer of 1 or more) sound collecting means (for example, microphones). The echo cancellation apparatus 100 includes N echo cancellation units 102 _n (n = 1,..., N) and an adaptive beam forming unit 104. In the following description, the reproduction signal emitted from the reproducing means 2 and x (t), and the reproduced signal x (t) is a signal that goes around the sound collecting means 4 _n echo signal a (t), the noise Let b (t) be the noise signal from the source and c (t) be the speaker signal from speaker 8. A signal consisting of an echo signal a (t), a noise signal b (t), and a speaker signal c (t) is used as a sound collection signal, and a signal collected by the sound collection means 4 _n is represented as y _n (t ), And the echo signal, noise signal, and speaker signal are indicated as a _n (t) b _n (t) c _n (t), respectively. t indicates a discrete time, which is indicated by sampling a continuous time signal at a constant sampling interval T. For example, the sampling frequency f _s = 48 kHz (48000 times per second). Sampling frequency f _s = 1 / T. Also, until the echo signal, noise signal, and speaker signal reach the sound collection means, they should be expressed in real time, not in discrete time, but for the sake of simplicity, all are shown as discrete time t. Since the echo canceller 100 and the echo cancellers 200 and 300 described below handle digital signals that are signals represented by discrete time t, the input analog signal is converted into a digital signal by a DA converter and output. It is necessary to convert the digital signal to be converted into an analog signal by the AD converter. However, since this is a matter of course, the AD converter and the DA converter are not shown in FIGS. When the noise source 6 and the speaker 8 are not present, the collected sound signal does not include a noise signal and a speaker signal.

Returning to FIG. The collected sound signal y _n (t) is input to the echo canceller 102 _n . Echo canceling unit 102 _n may suppress the echo signal _a n (t) from the sound collection signal _y n (t). Then, the residual echo signal e _n is a signal that the echo signal is erased _(t) is input to the adaptive beamforming unit 104.

Adaptive beamforming unit 104 of the N residual echo signal _{e n (t) (= 1} , ..., N) and suppressing noise signals from, and output as an output signal z (t). Detailed processing of the echo canceling unit 102 _n and the adaptive beamforming unit 104 will be described in “Best Mode for Carrying Out the Invention” below.

Configuration of the echo canceller 100 is set up adaptive beam forming section 104 in the subsequent stage of the N echo cancellation unit 102 _n. Therefore, the acoustic directivity of the variation due to the adaptive beamforming of the adaptive beam forming unit 104 so as not included in the echo path model of the echo canceling unit 102 _n.

Next, FIG. 2 shows a functional configuration example of a conventional echo canceling apparatus 200 equipped with adaptive beamforming. In the configuration of the echo canceller 200, the adaptive beamforming unit 202 is provided at the front stage and the echo canceler 204 is provided at the rear stage. With this configuration, when there is one output signal from the adaptive beamforming unit 202, the echo cancellation unit 204 performs one echo cancellation processing (adaptive filter processing) on the output signal. All that is required is a low computation and low memory configuration.
Patent No. 4104626 Kellermann, W., "Strategies for combining acoustic echo cancellation and adaptive beamforming microphone arrays," IEEE International Conference on Acoustics, Speech, and Signal Processing, vol.1, 21-24, pp.219-222, Apr.1997.

With the configuration of the echo canceling apparatus 100, it is necessary to provide the echo canceling unit 102 _n having a high calculation cost for the number N of the sound collecting means, which requires a huge calculation cost. Also, with the configuration of the echo canceller 200, time-varying beamforming must be taken into account in the echo path model of the echo canceler, so that if the adaptive beamforming unit 202 causes variations in acoustic directivity, The echo canceling unit 204 must re-estimate the echo path model (adaptive filter), and the desired echo canceling amount may not be reached.

  The object of the present invention is to (1) reduce the calculation cost, and (2) reduce the time required to reach the desired echo cancellation amount of the echo cancellation unit, thereby reducing the acoustic directivity due to adaptive beamforming processing. It is another object of the present invention to provide an echo canceling device that reaches a desired echo canceling amount.

  The echo canceller of the present invention cancels an echo signal included in a sound pickup signal picked up by N sound pickup means, suppresses a noise signal included in the sound pickup signal, and outputs it as an output signal. Is. The echo cancellation apparatus includes N low-frequency echo cancellation units, an adaptive beam forming unit, and an echo cancellation unit. The low-frequency echo canceling unit cancels a component of the echo signal included in a frequency band lower than a predetermined reference value of the collected sound signal, and outputs a residual echo signal. The adaptive beamforming unit suppresses the noise signal from the N residual echo signals and outputs an adaptive beamforming signal. The echo canceller outputs an output signal by canceling the echo signal from the adaptive beamforming signal.

  In the present invention, the desired echo cancellation amount in the low frequency band is referred to as the “high frequency” (hereinafter simply referred to as “high frequency”) because of the nature of the speaker voice in which power is concentrated in the low frequency band (hereinafter simply referred to as “low frequency”). Note that it is larger than the desired echo cancellation amount of. Since the N low-frequency echo canceling units cancel echoes only for a low frequency lower than a predetermined reference value of the input sound pickup signal, N echo canceling units in the echo canceling device 100 Computational cost can be reduced in comparison. Therefore, the problem of the echo canceller 100 can be solved.

  The echo canceling apparatus according to the present invention is configured in the order of N low-frequency echo canceling units, a low-frequency beam forming unit, and an echo canceling unit from the previous stage. Therefore, in the echo path model of the low-frequency echo canceling unit, re-estimation due to a change in acoustic directivity of the adaptive beamforming unit 104 can be avoided. In addition, for the echo canceler, the low-frequency echo canceler cancels the low-frequency echo, that is, cancels the echo signal components for the signal from which most of the echo signal components are deleted (adaptive beamforming signal). To do. Therefore, the power of the entire area residual echo signal used in the echo canceling unit can be reduced from the beginning, and as a result, the time required to reach the desired echo canceling amount of the echo canceling unit can be shortened. Therefore, even if the sound directivity fluctuation is caused by the adaptive beam forming unit, the echo canceling unit can stably achieve a desired echo canceling amount. Further, Non-Patent Document 1 does not describe the processing procedures of the echo cancellation technique and the adaptive beam forming technique.

  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. 3 shows a functional configuration example of the echo canceller 300, and FIG. 4 shows a processing flow. The echo cancellation apparatus 300 includes N low-frequency echo cancellation units 302 _n , an adaptive beam forming unit 104, and an echo cancellation unit 304. FIG. 5 shows a functional configuration example of the low-frequency echo cancellation unit 302 _n , FIG. 6 shows a functional configuration example of the adaptive beamforming unit 104, and FIG. 7 shows a functional configuration example of the echo cancellation unit 304. Note that the constituent means of all the low-frequency echo cancellers 302 _n (n = 1,..., N) are all the same, and therefore the suffix “n” is used for the reference numbers of the constituent means shown in FIG. Omitted. The echo canceling unit 304 is the same as the echo canceling unit 204 shown in FIG.

  As shown in FIG. 5, the low-frequency echo canceling unit 302 of this embodiment includes a low-pass filter unit 3014, a downsampling unit 3016, an adaptive filter unit 3018, a band dividing unit 3012, a subtracting unit 3020, and a band synthesis. And means 3022. The band dividing unit 3012 includes a low-pass filter unit 30122, a high-pass filter unit 30124, and two down-sampling units 30126. As shown in FIG. 6, the adaptive beamforming unit 104 of this embodiment includes a frequency domain conversion unit 1042, an adaptive beamforming unit 1044, a time domain conversion unit 1046, and an impulse response storage unit 1050 (or a steering vector). Storage unit 1048). Further, as shown in FIG. 7, the echo canceling unit 304 of this embodiment includes an adaptive filter unit 3048 and a subtracting unit 3051.

[Processing of low-frequency echo canceling unit 302 _n (step S102)]
First, the collected sound signal y _n (t) is input to the low-frequency echo canceling unit 302 _n . Echo canceling unit 302 _n erases the component of the echo signal _a n included in the frequency band lower than the reference value α of predetermined sound collecting signal _{y n (t) (t)} , the residual echo signal _e n (t ) Is output. First, the reference value α will be described. In the present invention, attention is paid to the fact that the low-frequency desired echo cancellation amount is larger than the high-frequency desired echo cancellation amount due to the nature of the speaker voice whose power is concentrated in the low frequency range. For example, if the total frequency band is 24 kHz, the speaker voice often concentrates below the reproduction signal frequency band of 12 kHz. As a result, if the component of the echo signal included in the sound collection signal frequency band of less than 12 kHz is eliminated. Therefore, most echo signal components can be eliminated. The reference value α indicates an upper limit frequency value of a band in which a signal that is a source of the echo signal (for example, a speaker voice signal) is concentrated. The reference value α is predetermined, and α = 12 kHz in the above example. As described above, the low-frequency echo canceling unit 302 _n performs the echo canceling process only for the low-frequency sound pickup signal y _n (t) in order to reduce the calculation cost. The reference value α of this frequency band may be set in advance. Hereinafter, the processing of the low-frequency echo canceling unit 302 will be specifically described with reference to FIG. For the collected sound signal y _n (t), the collected sound signal input to the low-frequency echo canceling unit 302 is indicated as “y (t)” for the sake of simplicity.

The collected sound signal y (t) collected and inputted by the sound collecting means is inputted to the low pass filter means 30122 and the high pass filter means 30124 in the band dividing means 3012. The low-pass filter unit 30122 convolves the sound pickup signal y (t) with a coefficient of a low-pass filter (in the above example, only a signal having a frequency of less than 12 kHz is passed) to obtain a low-frequency sound pickup signal y _L (t). Output.

On the other hand, the high-pass filter means 30124 convolves the sound pickup signal y (t) with the coefficient of the high-pass filter (in the above example, only the signal of 12 kHz or higher is passed) to obtain the high-frequency sound pickup signal y _H (t ) Is output. In the low-pass filter unit 30122 and the low-pass filter unit 3014, it is preferable that the reference values α of the frequency bands to be passed are the same.

The low-frequency sound collection signal y _L (t) and the high-frequency sound collection signal y _H (t) are input to the downsampling means 30126 and 30128, respectively. Then, the downsampling unit 30126 performs a downsampling process for thinning out, for example, M−1 pieces of the low frequency sound pickup signal y _L (t). The thinning-out number M is a positive real number equal to or greater than 1, for example, “2”. The signal after the downsampling process is represented as y _L (Mt). Similarly, the down-sampling means 30128 performs a down-sampling process for thinning out, for example, M−1 for the high frequency sound pickup signal y _H (t). The signal after the downsampling process is expressed as y _H (Mt). The low frequency sound collected signal y _L (Mt) is input to the subtracting means 3020, and the high frequency sound collected signal y _H (Mt) is input to the band synthesizing means 3022.

On the other hand, the reproduction signal x (t) is input to the reproduction unit 2 and the low-pass filter unit 3014. The low-pass filter means 3014 convolves the reproduction signal x (t) with the coefficient of the low-pass filter (in the above example, only the signal less than 12 kHz is passed), so that the low-frequency signal of the reproduction signal x (t) (Hereinafter referred to as “low frequency reproduction signal x _L (t)”). The low frequency reproduction signal x _L (t) is input to the downsampling means 3016. The down-sampling means 3016 performs M-1 down-sampling processing on the low frequency reproduction signal x _L (t). The signal x _L (Mt) after the downsampling process is input to the adaptive filter means 3018.

  Here, the downsampling processing by the downsampling means 3016, 30126, and 30128 may be lower than the original sampling frequency. The rational downsampling is described in detail in, for example, the up / down sampling apparatus and the up / down sampling apparatus of Japanese Patent Application No. 2007-262986 and the program thereof. Further, the downsampling means 3016, 30126, 30128 may be omitted.

ここで、Ｓはタップ数を示す。 The adaptive filter means 3018 outputs a low-frequency pseudo echo signal d _L (Mt) by convolving the low-frequency reproduction signal x _L (Mt) with the coefficient h _L (Mt) of the adaptive filter. The convolution formula is as follows.

Here, S indicates the number of taps.

ただし、
Ｘ_Ｌ（Ｍｔ）＝［ｘ_Ｌ（Ｍｔ）、ｘ_Ｌ（Ｍ（ｔ−１））、．．．、ｘ_Ｌ（Ｍ（ｔ−Ｓ＋１））］
Ｈ_Ｌ（Ｍｔ）＝［ｈ_Ｌ、０（Ｍｔ）、ｈ_Ｌ、１（Ｍｔ）、．．．、ｈ_{Ｌ、Ｓ−１}（Ｍｔ）］であり、βはスカラ量のステップサイズであって、取りうる範囲は０＜β＜２である。また、δは分母が０になることを防止するための微小な定数である。＜学習同定アルゴリズムの説明は以上＞ The subtracting means 3020 outputs a low-frequency residual echo signal e _L (Mt) by taking the difference between the low-frequency sound collection signal y _L (Mt) and the low-frequency pseudo echo signal d _L (Mt). In the example of FIG. 5, the low-frequency pseudo echo signal d _L (Mt) is subtracted from the low-frequency sound collection signal y _L (Mt). The low band residual echo signal e _L (Mt) is input to the band synthesizing unit 3022 and the adaptive filter unit 3018. The coefficient of the adaptive filter of the adaptive filter means 3018 is updated by the low frequency reproduction signal x _L (Mt) and the past low frequency residual echo signal e _L (Mt). Various algorithms are known for this update. For example, there are a learning identification algorithm (NLMS: Normalized Least-Mean-Squares) and an exponential weighting algorithm, but the learning identification algorithm will be briefly described just in case. The adaptive filter coefficient vector H _L (Mt) is updated by the following expression using the vector X _L (Mt) of the low frequency reproduction signal x _L (Mt) by the learning identification algorithm.

However,
X _L (Mt) = [x _L (Mt), x _L (M (t−1)),. . . , X _L (M (t−S + 1))]
H _L (Mt) = [h _{L, 0} (Mt), h _{L, 1} (Mt),. . . , H _{L, S-1} (Mt)], β is the step size of the scalar quantity, and the possible range is 0 <β <2. Further, δ is a minute constant for preventing the denominator from becoming zero. <End of description of learning identification algorithm>

The band synthesizing unit 3022 performs band synthesis on the low-frequency residual echo signal e _L (Mt) and the high-frequency sound collection signal y _H (Mt), thereby outputting a residual echo signal e (t). As a specific example of the processing details, M−1 zero values are inserted between samples for the low-frequency residual echo signal e _L (Mt) and the high-frequency sound collection signal y _H (Mt). The residual echo signal e (t) is output by adding the respective output signals convolved with a filter that removes the generated imaging (folded spectrum of the sampling frequency before upsampling). Then, the above processing for all of the low-frequency echo canceling section _{302 n (n = 1, ...} , N) is performed, the residual echo signal _e n (t) is output from each of the low-frequency echo canceling unit 302 _n Is done.

[Processing of Adaptive Beamforming Unit 104 (Step S104)]
All N residual echo signals e _n (t) (n = 1,..., N) are input to the adaptive beamforming unit 104. Adaptive beamforming section 104 suppresses the noise signal from N residual echo signals e _n (t) and outputs adaptive beamforming signal s (t). Hereinafter, a specific example will be described with reference to FIG. Details of the following description of the adaptive beamforming unit 104 are described in paragraphs [0004] to [0010] of Japanese Patent Application Laid-Open No. 2008-60635.

The input N residual echo signals e _n (t) are input to the frequency domain transforming means 1042. Frequency domain transform section 1042 converts the residual echo signal _e n (t) is a signal in the time domain frequency domain signal _E n (f, _τ) on. Here, f indicates a frequency, and τ indicates an arbitrary time. For the transform in the frequency domain, a known technique such as Fourier transform may be used. E _n (f, τ) (n = 1,..., N) are all input to adaptive beamforming means 1044.

The adaptive beamforming means 1044 emphasizes the speaker signal c (t) as the target signal and suppresses the noise signal b (t) as the unnecessary signal as much as possible (W ₁ (f),..., W _N. (F)] This is realized by estimating ^T. When designing the adaptive beamforming means 1044, the impulse response vector g (f) from the speaker to each sound collection means or g (f) It is assumed that the steering vector a (f) = [exp (−i2πfτ ₁ ),..., Exp (−i2πfτ _N )] ^T is known ”, where τ _n is from speaker 8. speaker signal (target signal) is the time difference reaches a time and origin 0 reaching the sound collection unit n. Then, it is often to linearly arranged a sound collecting means 4 _n as shown in FIG. speaker The direction of 8 is θ, and sound collecting s means 4 ₁ When coordinates _{d n} of the sound pickup means _{4 n} in the case of a reference (origin) of, tau _n described above is given by _{τ n = d n cosθ / c} . Where c is the signal rate of .

Returning to FIG. 6, as adaptive beamforming means 1044 for suppressing unnecessary signals, an adaptive filter group (vector) 1048 _n (n) that minimizes output power A (W (f)) expressed by the following equation. = 1,..., N) W (f) = [W ₁ (f),. . . , W _N (f)].
A (W (f)) = V {| S | ² (f, τ)}
= V {S (f, τ) S ^* (f, τ)}
= V {W ^T (f) E (f, τ) E ^T (f, τ) W (f)}
= W ^T (f) R _E (f) W (f) (1)

Here, V {} is an average operation with respect to time τ, A ^* is a complex conjugate of A, R _E (f) = V {E (f, τ) E ^T (f, τ)} is a correlation matrix of an input signal, S (f, τ) is an output of the adaptive beamforming means 1044 and can be expressed by the following equation.
S (f, τ) = W ^T (f) E (f, τ) (2)
Here, in order to avoid a meaningless solution (W (f) = 0 = [0,..., 0] ^T ), the constraint condition shown in the following equation is given that the target signal can be obtained without distortion. To do.
W ^T (f) g (f) = 1 (3)

Thus, the problem of obtaining the value of W (f) that satisfies Equation (3) and minimizes the value of A (W (f)) in Equation (1) is as follows using Lagrange's undetermined multiplier p. (4).
A ′ (W (f)) = A (W (f)) + p (W ^T (f) g (f) −1) (4)
By solving equation (4), the adaptive filter group (vector) W (f) is obtained by the following equation (5).

  The frequency domain adaptive beamforming signal S (f, τ) output from the adaptive beamforming means 1044 is input to the time domain transforming means 1046. The time domain transforming unit 1046 converts the adaptive beamforming signal S (f, τ), which is a frequency domain signal, into a time domain adaptive beamforming signal s (t) using, for example, a known technique such as inverse Fourier transform. Convert and output.

[Processing of Echo Erasing Unit 304 (Step S106)]
The adaptive beamforming signal s (t) is input to the echo canceller 304. The echo canceller 304 cancels the echo signal from the adaptive beamforming signal s (t), and outputs it as an output signal z (t). The processing content of the echo canceling unit 304 is known, but will be described with reference to FIG.

The reproduction signal x (t) is input to the adaptive filter means 3048. The adaptive filter means 3048 outputs a pseudo echo signal vector d (t) by convolving the reproduced signal x (t) with the coefficient h (t) of the adaptive filter as shown in the following equation.

  Then, the subtracting means 3051 outputs a residual echo signal e (t) by subtracting the pseudo echo signal d (t) from the adaptive beamforming signal s (t). Then, the vector of the adaptive filter of the adaptive filter means 3048 is updated by using the above-described learning identification method or the like with the past residual echo signal and the reproduced signal.

Here, the components of the echo signal included in the low frequency band of the collected sound signal are canceled by the N low-frequency echo canceling units 302 _n , that is, most of the components of the echo signal included in the adaptive beamforming signal are Is erased, the power of the residual echo signal e (t) becomes considerably small. Therefore, the time required to reach the desired echo cancellation amount can be shortened, and the desired echo cancellation amount can be reached even if the acoustic directivity of the adaptive beamforming unit 104 varies.

  The residual echo signal e (t) is output as an output signal z (t). When the speaker signal c (t) is present, the residual echo signal e (t) and the speaker signal c (t) are output as the output signal z (t).

The echo canceling unit 304 of the first embodiment deletes the components of the echo signal for the entire frequency band of the input signal. However, since the components of the low-frequency echo signal are canceled by the N low-frequency echo canceling units 302 _n , the echo canceling unit has a frequency band higher than the reference value α (α = 12 kHz in the above example). The echo signal may be deleted from the adaptive beamforming signal. By doing in this way, the calculation cost of an echo cancellation part can be reduced significantly. In this case, the reference numeral of the echo canceling unit is 306, and a functional configuration example of the echo canceling unit 306 is shown in FIG. The echo canceling unit 306 is configured by replacing the high-pass filter unit and the low-pass filter unit of the low-frequency echo canceling unit 302 (see FIG. 5). The processing contents of the echo canceling unit 306 will be briefly described.

First, adaptive beamforming signal s (t) is input to high-pass filter means 30622 and low-pass filter means 30624 in band dividing means 3062. The high-pass filter means 30622 convolves the adaptive beamforming s (t) with a coefficient of a high-pass filter (in the above example, only a signal having a frequency of 12 kHz (= reference value α) or higher) is passed, and high-frequency adaptive beamforming is performed. The signal s _H (t) is output. The low-pass filter unit 30624 convolves the adaptive beamforming signal s (t) with the coefficient of the low-pass filter (in the above example, only the signal less than 12 kHz is passed), and the low-pass adaptive beamforming signal s _L (t ) Is output.

The high-frequency adaptive beamforming signal s _H (t) and the low-frequency adaptive beamforming signal s _L (t) are input to downsampling means 30626 and 30628, respectively. Then, the downsampling means 30626 and 30628 respectively perform downsampling processing on the high frequency adaptive beamforming signal s _H (t) and the low frequency adaptive beamforming signal s _L (t). The high frequency adaptive beamforming signal s _H (Mt) after the downsampling process is input to the subtracting unit 3070, and the low frequency adaptive beamforming signal s _L (Mt) after the downsampling process is input to the band synthesizing unit 3072. The

On the other hand, the high-pass filter means 3064 applies the coefficient of the high-pass filter (in the above example, only the signal of 12 kHz (= reference value α) or higher) to the reproduction signal x (t) with respect to the reproduction signal x (t). Is output as a high-frequency reproduction signal x _H (t). The high frequency reproduction signal x _H (t) is input to the downsampling means 3066. The downsampling means 3066 performs a downsampling process on the high frequency reproduction signal x _H (t). The signal x _H (Mt) after the downsampling process is input to the adaptive filter means 3068. Further, the downsampling means 3066, 30626, and 30628 may be omitted.

The adaptive filter means 3068 outputs a high-frequency pseudo echo signal d _H (Mt) by convolving an adaptive filter with the low-frequency reproduction signal. The subtracting unit 3070 outputs a high-frequency residual echo signal e _H (Mt) by subtracting the high-frequency pseudo echo signal d _H (Mt) from the high-frequency adaptive beamforming signal s _H (Mt). The high-frequency residual echo signal e _H (Mt) is input to the band synthesizing unit 3072 and the adaptive filter unit 3068. The adaptive filter coefficient of the adaptive filter means 3068 is updated using the learning identification algorithm or the like by the high frequency reproduction signal x _H (Mt) and the past high frequency residual echo signal e _H (Mt).

The band synthesizing unit 3072 outputs a residual echo signal e (t) by performing band synthesis and upsampling processing on the high-frequency residual echo signal e _H (Mt) and the low-frequency adaptive beamforming signal s _L (Mt). To do. The details of the processing of the band synthesizing unit 3072 are the same as the processing of the band synthesizing unit 3022, and thus will be omitted. Then, the residual echo signal e (t) is output from the band synthesizing unit 3072 as the output signal z (t). <Description of processing of echo canceling unit 306>

The above-described echo canceling apparatus 300 has been described with respect to the case where there is one reproducing unit 2, but the present invention can be applied even when there are two or more reproducing units.
Further, an example has been described in which the processing of the low-frequency echo canceling unit 302 _n and the echo canceling unit 304 is performed in the time domain, and the processing of the adaptive beamforming unit 104 is performed in the frequency domain. However, the processing of the low-frequency echo cancellation unit 302 _n and the echo cancellation unit 304 may be performed in the frequency domain, and the processing of the adaptive beamforming unit 104 may be performed in the time domain.

In the present invention, attention is paid to the fact that the low-frequency desired echo cancellation amount is larger than the high-frequency desired echo cancellation amount due to the nature of the speaker voice whose power is concentrated in the low frequency range. The echo canceling apparatus according to the present embodiment is “low-frequency echo canceling processing by the low-frequency echo canceling unit 302 _n ” → “adaptive beamforming processing by the adaptive beamforming unit 104” → “echo canceling by the echo canceling unit 304”. Processing is performed in the order of. First, the low-frequency echo canceling unit 302 _n cancels the echo signal component included only in the low-frequency band of the collected sound signal. Therefore, the calculation cost can be reduced as compared with the echo canceller 100. In addition, low-frequency echo cancellation processing is performed before adaptive beamforming processing. Therefore, in the echo path model of the low-frequency echo canceling unit, re-estimation due to a change in acoustic directivity of the adaptive beamforming unit 104 can be avoided. Further, most of the echo signal components included in the collected sound signal are erased by the low-frequency echo cancellation processing. Accordingly, the power of the residual echo signal e (t) from the subtracting means 3051 in the echo canceling unit 304 becomes small, and as a result, the time required to reach the desired echo canceling amount of the echo canceling unit can be shortened. Therefore, even if the acoustic directivity characteristic varies due to the adaptive beam forming unit 104, a desired echo cancellation amount can be reached.

  An experimental result showing that the echo canceling apparatus 300 of the first embodiment is superior to the conventional echo canceling apparatus 100 will be described. When the number of microphones is N = 4, the entire frequency band is 24 kHz, the reference value α of the frequency band is 12 kHz, that is, the low band is less than 12 kHz and the high band is 12 to 24 kHz, the echo canceling apparatus 300 of this embodiment is used. This calculation cost can be reduced by about 40% from the calculation cost for the echo cancellation processing of the echo cancellation apparatus 100.

<Hardware configuration>
The present invention is not limited to the above-described embodiment. In addition, the various processes described above are not only executed in time series according to the description, but may be executed in parallel or individually according to the processing capability of the apparatus that executes the processes or as necessary. Needless to say, other modifications are possible without departing from the spirit of the present invention.
Further, when the above-described configuration is realized by a computer, processing contents of functions that the echo canceling apparatus 300 should have are described by a program. The processing function is realized on the computer by executing the program on the computer.

  The program describing the processing contents can be recorded on a computer-readable recording medium. The computer-readable recording medium may be any medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, or a semiconductor memory. Specifically, for example, the magnetic recording device may be a hard disk device or a flexible Discs, magnetic tapes, etc. as optical disks, DVD (Digital Versatile Disc), DVD-RAM (Random Access Memory), CD-ROM (Compact Disc Read Only Memory), CD-R (Recordable) / RW (ReWritable), etc. As the magneto-optical recording medium, MO (Magneto-Optical disc) or the like can be used, and as the semiconductor memory, EEP-ROM (Electronically Erasable and Programmable-Read Only Memory) or the like 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 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. Also, the program is not transferred from the server computer to the computer, and the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes the processing function only by the execution instruction and result acquisition. It is 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 present 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 echo canceller 300 described in this embodiment includes a CPU (Central Processing Unit), an input unit, an output unit, an auxiliary storage device, a RAM (Random Access Memory), a ROM (Read Only Memory), and a bus. (Both not shown).
The CPU executes various arithmetic processes according to the read various programs. The auxiliary storage device is, for example, a hard disk, an MO (Magneto-Optical disc), a semiconductor memory, or the like, and the RAM is an SRAM (Static Random Access Memory), a DRAM (Dynamic Random Access Memory), or the like. The bus connects the CPU, the input unit, the output unit, the auxiliary storage device, the RAM, and the ROM so that they can communicate with each other.

<Cooperation between hardware and software>
The word adding device of this embodiment is constructed by reading a predetermined program into the hardware as described above and executing it by the CPU. The functional configuration of each device constructed in this way will be described below.
The low-frequency echo canceling unit 302 _n , the adaptive beamforming unit 104, and the echo canceling unit 304 of the echo canceling apparatus 300 are arithmetic units constructed by reading a predetermined program into the CPU and executing it. A storage unit (not shown) of the echo canceller 300 functions as the auxiliary storage device.

The figure which showed the function structural example of the conventional echo cancellation apparatus. The figure which showed the function structural example of the conventional echo cancellation apparatus. FIG. 3 is a diagram illustrating a functional configuration example of an echo canceling apparatus according to the first embodiment. FIG. 3 is a diagram showing a processing flow of the echo canceling apparatus of the first embodiment. FIG. 3 is a diagram illustrating a functional configuration example of a low-frequency echo canceling unit according to the first embodiment. FIG. 3 is a diagram illustrating a functional configuration example of an adaptive beam forming unit according to the first embodiment. FIG. 3 is a diagram illustrating a functional configuration example of an echo canceling unit according to the first embodiment. The figure for demonstrating (tau) used by an adaptive beam forming process. FIG. 6 is a diagram illustrating a functional configuration example of an echo canceling unit according to the second embodiment.

Claims (8)

In an echo canceller that cancels an echo signal included in a sound pickup signal picked up by N sound pickup means, suppresses a noise signal included in the sound pickup signal, and outputs it as an output signal.
N low-frequency echo cancellers each canceling a component of an echo signal included in a frequency band lower than a predetermined reference value of the collected sound signal and outputting a residual echo signal;
An adaptive beamforming unit that suppresses the noise signal from N residual echo signals and outputs an adaptive beamforming signal;
An echo canceling unit that outputs the output signal by canceling an echo signal from the adaptive beamforming signal;
Echo canceller with The echo canceller according to claim 1,
The low-frequency echo canceller is
Low-pass filter means for outputting a signal in the low frequency band of the reproduction signal (hereinafter referred to as “low frequency reproduction signal”);
Band dividing means for dividing the sound collection signal into a low frequency sound collection signal and a high frequency sound collection signal by dividing the band,
Adaptive filter means for outputting a low-frequency pseudo echo signal by convolving an adaptive filter with the low-frequency reproduction signal;
Subtracting means for outputting a low-frequency residual echo signal by taking the difference between the low-frequency sound pickup signal and the low-frequency pseudo echo signal;
Band synthesis means for outputting the residual echo signal by performing band synthesis for the low-frequency residual echo signal and the high-frequency sound collection signal,
The adaptive filter is updated by using a past low-frequency residual echo signal and the low-frequency reproduction signal, and an echo canceller. The echo canceller according to claim 1 or 2, wherein
The echo canceller is
An echo canceller for canceling an echo signal from an adaptive beamforming signal in a frequency band higher than the reference value. In an echo canceling method of canceling an echo signal included in a sound pickup signal picked up by N sound pickup means, suppressing a noise signal included in the sound pickup signal, and outputting it as an output signal,
For each sound collection means, a low-frequency echo cancellation process for canceling the echo signal component included in a frequency band lower than a predetermined reference value of the sound collection signal and outputting a residual echo signal;
Adaptive beamforming process of suppressing the noise signal from the N residual echo signals and outputting an adaptive beamforming signal;
An echo cancellation process for outputting the output signal by canceling an echo signal from the adaptive beamforming signal;
An echo canceling method comprising: The echo cancellation method according to claim 4,
The low-frequency echo cancellation process includes:
A low-pass filter step of outputting a signal in a low frequency band of the reproduction signal (hereinafter referred to as “low frequency reproduction signal”);
By dividing the sound collection signal into bands, a band division step for outputting the sound collection signal divided into a low frequency sound collection signal and a high frequency sound collection signal,
An adaptive filter step of outputting a low-frequency pseudo echo signal by convolving an adaptive filter with the low-frequency reproduction signal;
A subtraction step of outputting a low-frequency residual echo signal by taking a difference between the low-frequency sound pickup signal and the low-frequency pseudo echo signal;
A band synthesis step of outputting the residual echo signal by performing band synthesis on the low-frequency residual echo signal and the high-frequency sound collection signal,
The adaptive filter is updated using a past low-frequency residual echo signal and the low-frequency reproduction signal, and an echo cancellation method characterized in that the adaptive filter is updated. The echo cancellation method according to claim 4 or 5,
The echo cancellation process includes:
An echo cancellation method comprising: canceling an echo signal from an adaptive beamforming signal in a frequency band higher than the reference value.   A program for operating a computer as the echo canceller according to claim 1. A computer-readable recording medium on which the program according to claim 7 is recorded.

Images