JP4156428B2 - Echo canceling method, echo canceling device, echo canceling program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an echo canceling apparatus for canceling echo that circulates from an acoustic reproduction means such as a speaker to an acoustic sound collection means such as a microphone.
[0002]
[Prior art]
The echo canceling device that cancels the echo that circulates from the sound reproducing means to the sound collecting means is connected as shown in FIG. Reference numeral 10 shown in FIG. 6 denotes a conventional echo canceling apparatus. In the conventional echo canceling apparatus 10, the impulse response h of the echo path between the sound reproducing means 1 and the sound collecting means 2 is estimated, and the estimated impulse response h ′ and the reproduced signal x input to the reproduced signal input terminal 3 are estimated. generating a convolution ^h'* x, it is subtracted from the actual echo signal y, to obtain a echo cancellation signal e to the echo cancellation signal output terminal 4.
However, the estimated impulse response and the reproduction signal convolution calculation require a large amount of calculation, which is a problem in implementation.
[0003]
In recent years, in order to solve this problem, a reproduction signal and a reverberation signal are once subjected to frequency domain transformation, a parameter corresponding to the frequency domain transformation of the impulse response of the echo path is estimated, and multiplication processing is used instead of convolution (non-convolution). Patent Document 1) or a method of reducing the amount of calculation has been proposed by dividing into smaller convolution operations (Patent Document 1). Examples of frequency domain transforms include (fast) discrete Fourier transform, (fast) discrete cosine transform, and (fast) Hartley transform.
In FIG. 6, a speaker is used as the sound reproducing unit 1, but the sound reproducing unit 1 includes an amplifier and a buffer in the previous stage of reproduction. Similarly, the sound pickup means 2 includes an amplifier and a buffer after the microphone.
[0004]
Moreover, the structure described in (nonpatent literature 1) etc. is typically shown in FIG. The configuration of FIG. 7 includes the following means. That is, the reproduction signal input means 101 that inputs the reproduction signal that is input from the reproduction signal input terminal 3 and is output to the sound reproduction means 1 and accumulates for a certain period of time to obtain a reproduction signal string, and the sound reproduction means 1 exist in the same space. A sound collection signal input means 102 for inputting a sound collection signal from the sound sound collection means 2 and storing it for a certain period of time to obtain a sound collection signal string; and a reproduction signal conversion means 103 for frequency domain conversion of the reproduction signal string to obtain a reproduction signal conversion string. Then, a reproduction signal conversion sequence is input for each frequency domain and filtered to generate a simulated reverberation signal conversion sequence that simulates the frequency domain conversion sequence of the reverberation signal that circulates from the sound reproduction means 1 to the sound collection means 2. Filter processing means 104, simulated error output means 105 for inputting a simulated echo signal conversion sequence and a collected sound signal sequence and outputting a simulated error of the simulated echo signal conversion sequence, reproduction signal conversion sequence and simulated error The filter error calculation means 106 that calculates the characteristic error of each filter domain of the filter coefficient of the filter processing means 104 is input, and the characteristic error calculated by the filter error calculation means 106 is multiplied by the first adjustment coefficient for each frequency domain. An update amount adjustment unit 107 and a filter update unit 108 that updates the filter coefficient of the filter processing unit 104 by adding a characteristic error multiplied by the first adjustment coefficient are included.
[0005]
[Patent Document 1]
JP-A-9-116472 (FIG. 6)
[Non-Patent Document 1]
Simon Haykin "Adaptive Filter Theory" Science and Technology Publishing, January 10, 2001, pp. 500-541
[0006]
[Problems to be solved by the invention]
In the configuration of FIG. 7, if the filter error calculation unit 106 can correctly calculate and output the characteristic error between the true characteristic for echo cancellation and the filter coefficient used in the filter processing unit 104, the filter error calculation unit 106 outputs the characteristic error. The first adjustment coefficient multiplied by the first update amount adjustment means 107 may be 1 for the characteristic error. That is, the characteristic error can be corrected most quickly.
However, it is difficult for the filter error calculation means 106 to actually calculate and output the characteristic error correctly. One of the causes is the influence of disturbance components such as ambient noise and circuit noise other than the echo signal included in the collected sound signal collected by the sound collecting means 2. This disturbance component adversely affects the calculation of the characteristic error output from the filter error calculation means 106. Therefore, in order to avoid this influence, the conventional technique is to introduce the first update amount adjusting means 107 and set the first adjustment coefficient small.
[0007]
Conventionally, since the first coefficient is set small, the speed at which the filter coefficient used in the filter processing unit 104 approaches the true characteristic for echo cancellation is low, and it takes a long time to achieve complete echo cancellation. It was.
As described above, the object of the present invention is to minimize the influence of disturbance components included in the collected sound signal picked up by the sound pickup means 2 and the first update adjustment means 107 takes the largest first adjustment coefficient. It is possible to increase the speed at which the filter coefficient used in the filter processing unit 104 approaches the true characteristic for echo cancellation.
[0008]
[Means for Solving the Problems]
According to the first aspect of the present invention, a reproduction signal input process for obtaining a reproduction signal sequence by inputting a reproduction signal to be output to the sound reproduction means and obtaining a reproduction signal string, and collecting from a sound collection means existing in the same space as the sound reproduction means. Sound signal input processing that inputs sound signals and accumulates them for a certain period of time to obtain a sound collection signal sequence, reproduction signal conversion processing that converts a reproduction signal sequence to a frequency domain and obtains a reproduction signal conversion sequence, and a reproduction signal conversion sequence for each frequency domain Filter processing for generating a simulated reverberation signal conversion sequence for simulating the frequency domain conversion sequence of the reverberation signal that circulates from the sound reproduction means to the sound collection means by being input to the filter, and the simulated reverberation signal conversion sequence and the sound collection signal And a simulated error output process for outputting a simulated error of the simulated echo signal conversion sequence, and a reproduction signal conversion sequence and the simulated error are input, and a characteristic error for each frequency domain of the filter coefficient of the filter processing is measured. A filter error calculation process, a first update amount adjustment process for multiplying the characteristic error calculated by the filter error calculation process by a first adjustment coefficient for each frequency domain, and a characteristic error multiplied by the first adjustment coefficient. In the echo cancellation method for executing the filter update process for updating the filter coefficient of the process, the magnitude of the disturbance signal component other than the echo signal included in the collected sound signal with respect to the magnitude of the simulated error output in the simulated error output process The ratio of the estimated value measured in advance is calculated as the disturbance ratio, and when the disturbance ratio deviates from the predetermined value range from 0, the disturbance ratio is set for each frequency domain by censoring the numerical value so that it falls within the range. Filter factor multiplied by the characteristic error of the filter coefficient or the first adjustment coefficient before or after the disturbance ratio estimation process for successive estimation and the first update amount adjustment process A second update amount adjustment process is performed by multiplying the characteristic error of 0 by a predetermined value between 0 and a predetermined value in accordance with the magnitude of the disturbance ratio, and by multiplying a second adjustment coefficient that takes a smaller value as the disturbance ratio increases for each frequency region. We propose an echo cancellation method.
[0009]
According to the second aspect of the present invention, the reproduction signal input means for inputting the reproduction signal output to the sound reproduction means and accumulating it for a predetermined time to obtain a reproduction signal string, and the sound collection means existing in the same space as the sound reproduction means are collected. A sound collection signal input means for inputting a sound signal and accumulating for a certain time to obtain a sound collection signal string, a reproduction signal conversion means for obtaining a reproduction signal conversion string by converting the reproduction signal string in a frequency domain, and a reproduction signal conversion string for each frequency domain Filter processing means for generating a simulated reverberation signal conversion sequence for simulating the frequency domain conversion sequence of the reverberation signal that circulates from the sound reproduction means to the sound collection means by being input to the filter, and a simulated reverberation signal conversion sequence and sound collection A simulated error output means for inputting a signal string and outputting a simulated error of the simulated echo signal conversion string, and a characteristic error for each frequency domain of the filter coefficient of the filter processing means by inputting the reproduction signal conversion string and the simulated error A filter error calculating means for calculating, a first update amount adjusting means for multiplying the characteristic error calculated by the filter error calculating means by a first adjustment coefficient for each frequency region, and a characteristic error multiplied by the first adjustment coefficient. In the echo canceller having the filter update means for updating the filter coefficient of the filter processing means, the magnitude of the disturbance signal component other than the echo signal included in the collected sound signal with respect to the magnitude of the simulated error output by the simulated error output means. The ratio of the estimated value measured in advance is calculated as the disturbance ratio, and when the disturbance ratio deviates from the predetermined value range from 0, the disturbance ratio is set for each frequency domain by censoring the numerical value so that it falls within the range. The filter multiplied by the characteristic error of the filter coefficient or the first adjustment coefficient in the previous stage or subsequent stage of the disturbance ratio estimation means for sequentially estimating and the first update amount adjustment means And a second update amount adjusting means for multiplying the characteristic error of the coefficient by a second adjustment coefficient that takes a smaller value as the disturbance ratio increases in accordance with the magnitude of the disturbance ratio between 0 and a predetermined value. An erasing device is proposed.
[0010]
According to a fifth aspect of the present invention, there is proposed an echo canceling program which is described by a code string readable by a computer and causes the computer to execute the echo canceling method according to the first or third aspect .
Disturbance component contained in the action collected sound signal sound pickup unit 2 is picked up according to the construction of this invention, in accordance with the ratio of the filter error calculation means had the characteristic error of the filter coefficients to be output, is calculated The uncertain part of the characteristic error of the filter coefficient can be reduced. For this reason, the first update amount adjusting means can take as large a first adjustment coefficient as possible, and the true speed at which the filter coefficient used in the filter processing means approaches for echo cancellation can be increased.
[0011]
In the disturbance ratio estimation means, the estimation of the disturbance ratio representing the ratio of the magnitude of the disturbance signal component other than the echo signal included in the sound pickup signal from the sound pickup means to the magnitude of the simulation error output by the simulation error output means is It can be realized as follows. Here, the magnitude of the simulated error output by the simulated error output means in the frequency domain ω (n = 1... N) divided into N is assumed to be | E (ω _n ) |, and the magnitude of the disturbance in the frequency domain ω _n is estimated. _{Let the} value be | N (ω _n ) |. As the disturbance estimated value | N (ω _n ) |, a value measured in advance in a silent section or the like is used. At this time, the disturbance ratio DR (ω _n ) is calculated as follows:
DR (ω _n ) = | N (ω _n ) | / | E (ω _n ) | (1)
Can be calculated as Here, the value of the disturbance ratio DR (ω _n ) is a value from 0 to a predetermined value T, for example, T = 1. If the value deviates from this range in an actual use environment, the range is from 0 to a predetermined value. Truncate the numerical value so that The value of DR (ω _n ) may be based on other than equation (1) as long as the ratio of | N (ω _n ) | to | E (ω _n ) |
[0012]
Further, the second adjustment coefficient β (ω _n ) of the second update amount adjusting means takes a value from 0 to a predetermined value, for example 1, and uses the above disturbance ratio DR (ω _n ),
β (ω _n ) = (T−DR (ω _n ) ^p ) ^{1 / p} (2)
Can be determined as follows. Here, p is a positive real number. Further, in addition to the expression (2), as shown in A of FIG. 4, the value of DR (ω _n ) is made to correspond smoothly, or as shown in FIG. Even if the value of DR (ω _n ) increases, the value of β (ω _n ) decreases as the value of DR (ω _n ) increases. Even when DR (ω _n ) is obtained in a form other than equation (1), an appropriate one can be selected from the correspondence between DR (ω _n ) and β (ω _n ) shown here.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 shows an embodiment of an echo canceling apparatus according to the present invention. The echo canceling apparatus shown in FIG. 1 will be described together with the echo canceling apparatus method.
Reference numeral 100 shown in FIG. 1 represents an echo canceling apparatus according to the present invention. In the echo canceling apparatus 100 according to the present invention, the following means are added to the conventional echo canceling apparatus 10 shown in FIG. That is, a disturbance ratio DR (ω _n ) representing the ratio of the magnitude of the disturbance signal component other than the echo signal included in the sound pickup signal from the sound pickup means 2 with respect to the magnitude of the simulation error output from the simulation error output means 105. The disturbance ratio estimating means 111 for successively estimating the disturbance ratio according to the magnitude of the disturbance ratio estimated by the disturbance ratio estimating means 111 to the characteristic error of the filter coefficient actually output by the filter error calculating means 106 The frequency of the second adjustment coefficient β (ω _n ), which is a predetermined value when 1 is 0%, for example, 1 when the disturbance ratio is 100%, and 0 when the disturbance ratio is between 0 and the predetermined value. Second update amount adjustment means 110 for multiplying each area.
[0014]
With this configuration, the disturbance coefficient included in the collected sound signal picked up by the sound pickup means 2 has the calculated filter coefficient according to the ratio of the characteristic error of the filter coefficient output from the filter error calculation means 106. The value of the uncertain part of the characteristic error can be reduced. For this reason, the first update amount adjusting means 107 can take as large a first adjustment coefficient as possible, and the speed at which the filter coefficient used in the filter processing means 104 approaches the true characteristic for echo cancellation can be increased.
[0015]
Modified embodiment of the invention In the embodiment shown in Fig. 1, the second update amount adjusting means 110 is provided in the preceding stage of the first update amount adjusting means 107, but as shown in Fig. 2, Even if the two update amount adjusting means 110 is arranged at the subsequent stage of the first update amount adjusting means 107, the same effect as that of the embodiment shown in FIG. 1 can be obtained.
[0016]
Further, as shown in FIG. 3, a third update amount adjustment unit 112 is arranged between the filter error calculation unit 106 and the filter update unit 108, and the first adjustment coefficient and the second adjustment are adjusted by the third update amount adjustment unit 112. For example, a product with a coefficient is obtained, the value of this product is set as a third adjustment coefficient, and the characteristic of the filter coefficient for each frequency domain is multiplied by the same function as that of the embodiment shown in FIG. An effect can be obtained.
The echo canceling apparatus and the echo canceling method according to the present invention described above are realized by causing a computer to execute a sound canceling program described by a computer-readable code string. FIG. 5 shows a state in which the sound erasing program according to the present invention is installed in the computer 20.
[0017]
As is well known, the computer 20 has a central processing unit (CPU) 21, a read-only memory (ROM) 22 in which basic programs for controlling the startup and shutdown of the computer, etc. are recorded, In addition to the temporary recording, the read / write memory (RAM) 23 storing the program for executing the echo canceling method of the present invention, the input port 24, the output port 25, and the like can be used.
The reproduction signal input terminal 3 and the sound pickup means 2 are connected to the input port 24, and the reproduction signal X and the sound collection signal are input thereto. Further, the sound reproducing means 1 and the echo canceling signal output terminal 4 are connected to the output port 25, and the sound corresponding to the reproducing signal and the echo canceling signal e = yy ′ are output.
[0018]
In addition to the data storage area 23A, the readable / writable memory 23 includes a reproduction signal input processing program 23B that operates as the reproduction signal input means 101, and a sound collection signal input processing program 23C that operates as the sound collection signal input means 102. A reproduction signal conversion processing program 23D operating as the reproduction signal conversion means 103, a filter processing program 23E operating as the filter processing means 104, a simulation error output processing program 23F operating as the simulation error output means 105, and a filter error calculation means A filter error calculation processing program 23G that operates as 106, a first update amount adjustment processing program 23H that operates as first update amount adjustment means 107, a filter update processing program 23I that operates as filter update means 108, and a second update amount Adjusting hand The second update amount adjustment processing program 23J that operates as 110 and the disturbance ratio estimation processing program 23K that operates as the disturbance ratio estimation means 111 are installed, and each of these programs is decoded by the central processing unit 21 to perform echo cancellation operation Execute. The programs 23B to 23K installed in the readable and writable memory 23 are recorded in advance on a magnetic disk or CD-ROM that can be read by a computer, and installed from these recording media or installed through a communication circuit. The
[0019]
【The invention's effect】
The echo canceling apparatus according to the present invention suppresses a decrease in the speed at which the filter coefficient used in the filter processing means approaches the true characteristic for echo canceling, while reducing the disturbance component included in the collected sound signal collected by the acoustic sound collecting means. The influence can be reduced and the echo canceling performance can be enhanced.
[Brief description of the drawings]
FIG. 1 is a block diagram for explaining an embodiment of an echo canceling apparatus according to the present invention;
FIG. 2 is a block diagram for explaining another embodiment of the echo canceling apparatus according to the present invention;
FIG. 3 is a block diagram for explaining still another embodiment of the echo canceling apparatus according to the present invention.
FIG. 4 is a characteristic curve diagram for explaining the operation of the main part of the present invention.
FIG. 5 is a conceptual diagram for explaining the configuration when the echo canceling apparatus according to the present invention is realized by a computer.
FIG. 6 is a block diagram for explaining the outline of the echo canceling apparatus.
FIG. 7 is a block diagram for explaining a conventional echo canceling apparatus.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Sound reproduction | regeneration means 105 Simulated error output means 2 Sound pickup means 106 Filter error calculation means 3 Playback signal input terminal 107 1st update amount adjustment means 4 Echo cancellation signal output terminal 108 Filter update means 100 Echo cancellation apparatus 110 2nd update amount Adjustment means 101 Reproduction signal input means 111 Disturbance ratio estimation means 102 Sound pickup signal input means 103 Reproduction signal conversion means 104 Filter processing means

Claims (5)

A reproduction signal input process for inputting a reproduction signal output to the sound reproduction means and accumulating a reproduction signal sequence for a certain period of time, and a sound collection signal inputted from the sound collection means existing in the same space as the sound reproduction means A sound collection signal input process for accumulating time to obtain a sound collection signal string, a reproduction signal conversion process for obtaining a reproduction signal conversion string by converting the reproduction signal string in a frequency domain, and a filter for inputting the reproduction signal conversion string for each frequency domain Filter processing for generating a simulated echo signal conversion sequence that simulates a frequency domain conversion sequence of an echo signal that circulates from the acoustic reproduction means to the acoustic sound collection means by processing, the simulated echo signal conversion sequence, and the collected sound signal A simulated error output process for inputting a sequence and outputting a simulated error of the simulated echo signal conversion sequence; and for each frequency domain of the filter coefficient of the filter processing by inputting the reproduction signal conversion sequence and the simulated error A filter error calculation process for calculating a characteristic error; a first update amount adjustment process for multiplying the characteristic error calculated by the filter error calculation process by a first adjustment coefficient for each frequency domain; and a characteristic multiplied by the first adjustment coefficient. In an echo canceling method for executing a filter update process for updating a filter coefficient of the filter process by adding an error,
Calculating a ratio of a pre-measured estimated value of the magnitude of a disturbance signal component other than the echo signal included in the collected sound signal to the magnitude of the simulated error output in the simulated error output process, as a disturbance ratio, When the disturbance ratio deviates from a predetermined value range from 0, a disturbance ratio estimation process for sequentially estimating the disturbance ratio for each frequency domain by performing numerical truncation so that the disturbance ratio falls within the range, and the first update Before or after the amount adjustment processing, the filter coefficient characteristic error or the filter coefficient characteristic error multiplied by the first adjustment coefficient is a disturbance between 0 and a predetermined value depending on the magnitude of the disturbance ratio. And a second update amount adjustment process in which a second adjustment coefficient that takes a smaller value as the ratio increases is multiplied for each frequency region. A reproduction signal input means for inputting a reproduction signal output to the sound reproduction means and accumulating for a certain period of time to obtain a reproduction signal sequence; and a sound collection signal inputted from the sound collection means existing in the same space as the sound reproduction means and constant Sound collection signal input means for accumulating time and obtaining a sound collection signal string, reproduction signal conversion means for obtaining a reproduction signal conversion string by frequency domain conversion of the reproduction signal string, and inputting and reproducing the reproduction signal conversion string for each frequency domain Filter processing means for generating a simulated reverberation signal conversion sequence for simulating a frequency domain conversion sequence of an echo signal that circulates from the sound reproduction means to the sound collection means by processing, the simulated reverberation signal conversion sequence, and the sound collection A simulated error output means for inputting a signal string and outputting a simulated error of the simulated echo signal conversion string; a frequency of a filter coefficient of the filter processing means for inputting the reproduction signal conversion string and the simulated error; A filter error calculating means for calculating a characteristic error for each region; a first update adjusting means for multiplying the characteristic error calculated by the filter error calculating means for each frequency region; and the first adjustment coefficient. In the echo canceller having the filter update means for updating the filter coefficient of the filter processing means by adding the characteristic error,
Calculating a ratio of a pre-measured estimated value of the magnitude of a disturbance signal component other than the echo signal included in the collected sound signal to the magnitude of the simulated error output by the simulated error output means, as the disturbance ratio, When the disturbance ratio deviates from a predetermined value range from 0, a disturbance ratio estimation unit that sequentially estimates the disturbance ratio for each frequency region by performing numerical truncation so that the disturbance ratio falls within the range, and the first update amount Before or after the adjustment means, a disturbance ratio is determined according to the magnitude of the disturbance ratio between 0 and a predetermined value to the filter coefficient characteristic error or the filter coefficient characteristic error multiplied by the first adjustment coefficient. A reverberation canceling device comprising: a second update amount adjusting unit that multiplies a second adjustment coefficient that takes a smaller value as the value increases for each frequency region. The echo cancellation method according to claim 1,
Each said frequency domain is ω _n (N = 1, 2,..., N) and the second adjustment coefficient is β (ω _n ), The predetermined value is T, and the disturbance ratio is DR (ω _n ), The second adjustment coefficient β (ω _n )
      β (ω _n ) = (T-DR (ω _n ) ^p ) ^{1 / p}       (Where p is a positive real number)
The echo canceling method characterized by calculating | requiring by. The echo canceling device according to claim 2,
Each said frequency domain is ω _n (N = 1, 2,..., N) and the second adjustment coefficient is β (ω _n ), The predetermined value is T, and the disturbance ratio is DR (ω _n ), The second adjustment coefficient β (ω _n )
      β (ω _n ) = (T-DR (ω _n ) ^p ) ^{1 / p}       (Where p is a positive real number)
An echo canceling device characterized by the above. An echo canceling program, which is described by a computer-readable code string, and causes the computer to execute the echo canceling method according to claim 1 or 3.

Images