JP4317526B2 - Acoustic echo cancellation method, apparatus, program, and recording medium - Google Patents

Description

  The present invention relates to a method and apparatus for erasing acoustic echoes, which sometimes cause howling, in a loudspeaker communication system, a program for realizing the apparatus with a computer, and a recording medium on which the program is recorded.

In a loudspeaker communication system, a received voice signal is loudened from a speaker and picked up by a microphone to generate an acoustic echo, and the processing thereof becomes a problem. Acoustic echo causes adverse effects such as telephone call disturbance and discomfort. If the closed-loop gain formed including the ground-based loudspeaker call system is greater than 1, the acoustic echo causes howling and makes the call impossible.
In order to realize a more natural voice call environment, an acoustic echo canceller including an adaptive filter is required to eliminate acoustic wraparound from the speaker to the microphone. FIG. 10 shows an acoustic echo canceller including a one-channel reproduction system and a one-channel sound collection system. The reception signal from the reception terminal 1 is reproduced as an acoustic signal by the speaker 2 that functions as a loudspeaker, and passes through the acoustic echo path to the microphone 3 that functions as a sound collection unit. In the echo cancel unit 4, the received signal x (k) is input to the FIR filter 41 for generating a pseudo echo signal, and a pseudo echo signal is generated. Echo cancellation is performed by subtracting the pseudo echo signal from the collected sound signal y (k) by the subtracting means 43. Based on the difference signal between the collected sound signal y (k) and the pseudo echo signal and the past received signal x (k), the filter coefficient is updated so that the difference between the collected sound signal y (k) and the pseudo echo signal is reduced. The filter coefficient of the FIR filter 41 is updated by the means 42. As an update method of the filter coefficient of the filter coefficient update means 42, for example, there is an update method described in Patent Document 1. The FIR filter 41 and the subtracting means 43 having variable filter coefficients constitute an adaptive filter.

The adaptive filter normally used in acoustic echo cancellers is based on a linear FIR filter. The path from the sound wave emitted from the speaker to the microphone is well modeled by a linear FIR filter.
Japanese Patent No. 2533353

However, the entire echo path to be estimated by the adaptive filter cannot always be satisfactorily modeled by the FIR filter, and sometimes a nonlinear component is mixed in the echo signal. The echo path to be estimated includes characteristics of the microphone, the microphone amplifier 8, and the AD converter 9. For example, when a signal having an excessive level compared to the rated input level is input, a nonlinear component is mixed in the echo signal.
When the echo signal includes a nonlinear component, the filter coefficient of the adaptive filter is erroneously updated, and the estimated value of the echo path is degraded. As an example, when an excessively high level collected sound signal is clipped by the AD converter 9, the adaptive filter determines that the impulse response of the entire echo path has changed, and the difference between the collected sound signal and the pseudo echo signal is zero. As a result, all the coefficients of the FIR filter 41 are corrected. As a result, the echo canceling function is deteriorated, and if it is worsened, howling or unpleasant noise may occur.

  An object of the present invention is to provide an acoustic echo cancellation method and apparatus for preventing an adaptive filter coefficient from being erroneously updated when a nonlinear component is included in an echo signal.

Therefore, in the present invention, whether or not a nonlinear component is included in the echo signal is determined from the microphone sound pickup signal, and the operation of the adaptive filter is controlled. When it is determined that the nonlinear component is not included in the echo signal, the filter coefficient of the adaptive filter is updated, and when it is determined that the nonlinear component is included, the filter coefficient of the adaptive filter is not updated. The algorithm is the gist of the present invention.
More specifically, the acoustic echo canceller proposed in Embodiment 1 of the present invention includes a loudspeaker that amplifies a received signal, a sound collector that collects a transmitted voice, an adaptive filter that generates a pseudo echo signal from the received signal, and Subtracting means for obtaining a difference signal from the pseudo echo signal from the collected sound signal to be a transmission signal, filter coefficient updating means for updating a filter coefficient used in the adaptive filter by receiving the received signal and the difference signal, When it is determined whether a non-linear component is included in the sound signal, and when it is determined that the non-linear component is not included, the update operation of the filter coefficient updating unit is executed, and it is determined that the non-linear component is included And non-linearity determining means for stopping the updating operation of the filter coefficient updating means.

  In Embodiment 2 of the present invention, in the acoustic echo canceller proposed in Embodiment 1, nonlinear processing is performed on the first subtracting means that obtains a difference signal from a pseudo echo signal from a collected sound signal and uses it as a transmission signal. When a second subtraction means for obtaining a difference signal from the pseudo-echo signal subjected to the non-linear processing from the sound pickup signal to obtain a transmission signal is provided, and the non-linearity determination means determines that there is no non-linear component in the sound pickup signal, Switching means using the difference signal output from the first subtracting means as the transmission signal and using the difference signal output from the second subtracting means as the transmission signal when the nonlinearity determining means determines that the collected sound signal has a nonlinear component. Is provided.

In Embodiment 3 of the present invention, in any of the acoustic echo cancellers proposed in Embodiment 1 or 2, the nonlinearity determination means compares the absolute value of the collected sound signal with a preset threshold value, A determination unit that determines that a nonlinear component is included when the absolute value is larger than the threshold value is provided.
In Embodiment 4 of the present invention, a loudspeaker that amplifies the received signal, a sound pickup unit that collects the transmitted voice, a received signal frequency band dividing unit that divides the received signal into subband received signals of a plurality of frequency bands, , A collected sound signal frequency band dividing means for dividing the collected sound signal into sub-band collected signals of a plurality of frequency bands, and a subband divided by each of the received signal frequency band dividing means and the collected sound signal frequency band dividing means A filter by an adaptive filter that receives a reception signal and a subband sound pickup signal and generates a subband pseudo echo signal from each subband reception signal, and a subband difference signal between the subband sound pickup signal and the subband pseudo echo signal Subtracting means for obtaining a subband received voice signal and a subband difference signal as inputs and a filter unit for updating a filter coefficient used in an adaptive filter The update means and whether or not the non-linear component is included in the sub-band sound pickup signal, and if it is determined that the non-linear component is not included, the operation of the filter coefficient update means is executed, and the non-linear component is included. When the determination is made, the non-linearity determination process for stopping the operation of the filter coefficient updating unit is executed for each subband, and a subband echo canceling unit for outputting a subband difference signal from each subband, and a subband for each subband. Frequency band synthesizing means for synthesizing the sub-band difference signal output after the band echo cancellation processing to produce a transmitted voice signal.

  In the fifth embodiment of the present invention, in the acoustic echo canceling apparatus described in the fourth embodiment, the subtracting means provided in the subband echo canceling unit is the first subtracting means, and the first subtracting means is subjected to nonlinear processing. A second subtracting means for obtaining a subband difference signal from the band collected sound signal and a subband pseudo echo signal subjected to nonlinear processing to obtain a transmission signal is provided. In the nonlinearity determining means, a nonlinear component is included in the subband collected sound signal. When it is determined that the subband difference signal output from the first subtracting means is synthesized by the frequency band synthesizing means to be a transmission signal, the non-linearity judging means has a non-linear component in the subband sound pickup signal. Is determined, the subband difference signal output from the second subtracting means is synthesized by the frequency band synthesizing means to obtain a transmission signal.

  The present invention determines whether or not a non-linear component is included in the echo signal in the loudspeaker communication system, and when it is determined that the non-linear component is not included in the echo signal, the filter coefficient of the adaptive filter is updated, When it is determined that the nonlinear component is included, the updating of the filter coefficient of the adaptive filter is stopped. Thereby, when a nonlinear component is included in the echo signal, it is possible to prevent the adaptive filter coefficient from being updated erroneously, and it is possible to control the deterioration of the echo canceling function due to the mixing of the nonlinear component.

The sound echo canceling method and apparatus according to the present invention can be realized entirely by hardware. Preferably, the computer is provided with a sound echo canceling program proposed by the present invention installed in a computer. An embodiment in which a program is decoded by a CPU (central processing unit) and a computer functions as an acoustic echo canceller is the best embodiment.
When the computer functions as an acoustic echo canceller, the computer includes at least an adaptive filter that generates a pseudo echo signal from the received signal, and a subtracting unit that obtains a difference signal between the pseudo echo signal from the collected sound signal and uses it as a transmitted signal. A filter coefficient updating means for inputting a received signal and a difference signal to update a filter coefficient used for an adaptive filter, and determining whether or not a non-linear component is included in the collected sound signal. If it is determined, the filter coefficient update means performs the filter coefficient update operation, and if it is determined that a nonlinear component is included, constructs a nonlinearity determination means that stops the filter coefficient update means update operation. Each of these components interacts to function as an acoustic echo canceller.

A first embodiment of an acoustic echo canceling apparatus according to the present invention will be described below with reference to FIG. Portions corresponding to those of the configuration of the prior art shown in FIG. 10 are denoted by the same reference numerals, and redundant description thereof is omitted, but in the embodiment shown in FIG. 1, a loudspeaker call comprising a one-channel regenerator and a one-channel sound collector The case where this invention is applied to an apparatus is shown. A feature of the present invention is that the echo canceling unit 4 is provided with a nonlinearity determining unit 45.
The reception signal from the reception terminal 1 is reproduced as an acoustic signal by the speaker 2 and goes around the microphone 3 through an acoustic echo path. In the echo cancel unit 4, the received signal x (k) is input to the FIR filter 41 for generating a pseudo echo signal, and a pseudo echo signal is generated. By subtracting the pseudo echo signal from the collected sound signal by the subtracting means 43, the echo is canceled and transmitted from the transmission terminal 5 to the ground.

  The collected sound signal y (k) that has passed through the microphone amplifier 8 and the AD converter 9 is input to the nonlinearity determination unit 45, and it is determined whether the collected sound signal y (k) includes a nonlinear echo component. As one of the determination methods, a method of comparing the absolute value of the collected sound signal y (k) with a preset threshold value can be considered. For example, if this threshold value is set in the vicinity of the upper limit of AD conversion, it can be determined whether or not the collected sound signal y (k) is clipped by the AD converter 9. As shown in FIG. 8, if the level ani_p1 at which the input / output characteristic A of the microphone amplifier 8 becomes non-linear is measured in advance and the threshold is set to this level ani_p1, the collected sound signal y (k) is nonlinearly generated by the microphone amplifier 8. It can be determined whether or not an echo component is mixed.

  When the nonlinearity determination unit 45 determines that the nonlinear component is included in the echo signal, the update operation of the filter coefficient update unit 42 is stopped and the filter coefficient of the FIR filter 41 is not updated. Further, when it is determined that the nonlinear component is not included in the echo signal, the update operation of the filter coefficient updating means 42 is maintained, and the filter coefficient of the FIR filter 41 is set so that the difference between the collected sound signal and the pseudo echo signal becomes small. Update. As this update method, the NLMS algorithm, projection algorithm, ES algorithm, etc. introduced in Non-Patent Document 1 (Oga, Yamazaki, Kanada: Acoustic System and Digital Processing, IEICE 1995) can be used.

By stopping the update of the filter coefficient of the FIR filter 41 when it is determined that the nonlinear echo component is included in the echo signal, the filter coefficient of the adaptive filter is prevented from being erroneously updated by the nonlinear echo. . FIG. 2 shows a flowchart for explaining the outline of the acoustic echo cancellation program when this embodiment is realized by a computer.
In step SP1, the received signal is input to the FIR filter for generating a pseudo echo signal, and a pseudo echo signal is generated.
At step SP2, the pseudo echo signal is subtracted from the collected sound signal and transmitted to the ground.

In step SP3, it is determined whether a non-linear echo component is included in the collected sound signal picked up by the microphone.
If there is a nonlinear echo component in step SP4, the process branches to step SP5, and the filter coefficient updating operation is stopped. If there is no nonlinear echo component, the process branches to step SP6, and the filter coefficient updating operation is continued. When step SP5 or SP6 is executed, the process returns to step SP1 again, and the operations of steps SP1 to SP5 or SP6 are repeated.

  Next, Embodiment 2 of the present invention is shown in FIG. In this embodiment, the subtracting means 43 shown in the echo canceling section 4 described with reference to FIG. 1 is used as a first subtracting means, a second subtracting means 47 is added to the first subtracting means 43, and the first subtracting means 43 In this example, two echo cancellation methods are prepared using the second subtracting means 47. When it is determined by the nonlinearity determination unit 45 that the echo signal does not contain a nonlinear component, echo cancellation processing is performed by the first subtracting means 43 as in the case of FIG. When it is determined that the echo signal contains a non-linear component, the second subtracting means 47 performs non-linear processing similar to the pseudo echo signal and the collected sound signal, and the echo is eliminated. 48A and 48B constitute switching means for switching between two echo cancellation methods.

In this way, two types of echo cancellation processing are provided, and by switching the echo cancellation processing according to the nonlinearity determination result, erroneous update of the adaptive filter due to the nonlinear echo component can be prevented, and at the same time, the nonlinear echo component can also be canceled. Become.
That is, the non-linear processing means 44 is provided before the second subtracting means 47, and the non-linear processing means 46 is provided between the FIR filter 41 and the second subtracting means 47, so that the characteristics of the microphone amplifier 8 and the AD conversion unit 9 can be improved. The nonlinear echo component can be surely eliminated regardless of variations. As an input characteristic from the input to the microphone amplifier 8 to the output of the AD conversion unit 9, for example, a characteristic A as shown in FIG. It is linear when the input is smaller than ain_p1 (linear), and is saturated when it is larger than ain_p1 (nonlinear). The shape of the saturation curve that determines the nonlinear characteristics and the value of ain_p1 depend on the elements actually used in the electronic circuit and vary from device to device. Therefore, by performing non-linear processing on the digital signal after AD conversion by the non-linear processing means 44, characteristics from the input to the microphone amplifier 8 to the output of the non-linear processing means 44 are suppressed, for example, while suppressing variations in the non-linear characteristics for each apparatus. Limiter characteristics B like 8 can be obtained. At the same time, by setting the characteristics of the non-linear processing means 46 so as to coincide with the characteristics from the input to the microphone amplifier 8 to the non-linear processing means 44, the signals applied to the limiter characteristics B are approximated to each other, and the non-linear echo component is reliably ensured. It can be erased.

  Further, by setting the characteristics from the input to the microphone amplifier 8 to the output of the nonlinear processing means 46 and the characteristics of the nonlinear processing means 46 in, for example, a soft limiter as shown by a curve C in FIG. Degradation can be reduced.

Next, a third embodiment of the present invention is shown in FIG. In the third embodiment, the received signal x (k) is divided into D bands by the received signal band dividing unit 10, and the collected sound signal y (k) is divided into D bands by the collected signal band dividing unit 11. Then, echo cancellation processing is performed for each band by the subband echo cancellation units 4 ₁ to 4 _D. For the internal processing of the subband echo canceling units 4 ₁ to 4 _D , the processing of the first embodiment or the processing of the second embodiment is used. The subband sound pickup signal that has undergone the echo cancellation processing is synthesized by the transmission signal band synthesis unit 12 and transmitted from the transmission end 5 as transmission speech.

  By dividing the received signal x (k) and collected sound signal y (k) into bands and performing echo cancellation processing for each band, the echo path can be estimated quickly even when a colored signal such as an audio signal is input. This makes it possible to quickly erase the echo.

Next, Embodiment 4 of the present invention is shown in FIG. In the fourth embodiment, the M channel signal is reproduced by _M regenerators, that is, speakers 2 ₁ to 2 _{M, and the} N channel signal is collected by N sound collectors, that is, microphones 3 _{1 to} 3 _N. . M-channel echo canceling units 4 ₁ to 4 _N are arranged in the subsequent stages of the microphones 3 _{1 to} 3 _N , respectively, and _N M-channel reproduction 1-channel sound pickup systems are arranged in parallel. .
An example of the configuration of the M channel echo cancellation units 4 ₁ to 4 _N is shown in FIG. In FIG. 6, the transmission of the outputs of the FIR filters 41 _{1 to} 41 _M corresponding to the respective echo paths between the speakers 2 _{1 to} 2 _M and the microphone 3 is obtained by the adding means 49, and the sum is obtained by the subtracting means 43. Echo cancellation is performed by subtracting from.

When the nonlinearity determination unit 45 determines that the echo signal does not contain a nonlinear component, the filter coefficient update units 42 _{1 to} 42 _M continue the filter coefficient update operation, and the FIR filters 41 _{1 to} 41 _M Are updated so that the difference between the collected sound signal and the pseudo echo signal becomes small. As this updating method, the adaptive algorithm described in Patent Document 2 can be used. When it is determined that the nonlinear component is included in the echo signal, the filter coefficient update operation of the filter coefficient updating means 42 _{1 to} 42 _M is stopped, and the filter coefficients of the FIR filters 41 _{1 to} 41 _M are not updated. The configuration of FIG. 6 can be easily combined with the adaptive algorithm described in Patent Document 3. (Patent Document 2 Patent No. 3397269) (Patent Document 3 JP 2002-223182 A).

As described above, by providing the non-linearity determining unit 45 in each of the M channel echo canceling units 4 ₁ to 4 _N , the adaptive filter coefficient is erroneously updated when a non-linear component is included in the echo signal even in the multi-channel echo canceller. Can be prevented.

Next, another configuration example of 4 ₁ to 4 _N of the M channel echo cancellation unit is shown in FIG. In the fifth embodiment, as in the fourth embodiment, the M channel signal is reproduced by the _M speakers 2 ₁ to 2 _{M, the} N channel signals are collected by the N microphones 3 _{1 to} 3 _N , and the microphones 3 ₁ to 3 are collected. M channel echo canceling units 4 ₁ to 4 _N are respectively arranged after 3 _N.
In the fifth embodiment, the second subtracting means 47 is added to the echo canceling units 4 ₁ to 4 _N shown in FIG. 6, and two echo canceling methods are prepared. When the non-linearity determining unit 45 determines that the echo signal does not contain a non-linear component, the first subtracting unit 43 performs echo cancellation processing. When it is determined that a nonlinear component is included in the echo signal, the nonlinear processing means 46 similar to the sound collection signal is applied to the pseudo echo signal, and the echo is eliminated by the subtraction operation of the second subtraction means 47.

Thus, by providing two types of echo cancellation processing and switching according to the determination result of the non-linearity, it is possible to prevent erroneous update of the adaptive filter due to the non-linear echo component and simultaneously cancel the non-linear echo component. 5 to 7, the DA converter 6 and the amplifier 7 provided in the reproduction system, and the amplifier 8 and the AD converter 9 provided in the sound collection system are omitted.
The acoustic echo canceller according to the present invention described above can be configured by hardware, but the embodiment in which the acoustic echo canceler program is installed in the computer and the computer functions as the acoustic echo canceller is the simplest than that. This is a realization method. The present invention also claims an acoustic echo cancellation program installed on a computer. The acoustic echo canceling program according to the present invention is described in a computer-readable program language. The operating procedure described in the program language is decoded by a CPU provided in the computer, and the computer is explained with reference to FIG. 3 to 7 function as the acoustic echo canceling device described with reference to FIGS. The acoustic echo cancellation program is recorded on a computer-readable recording medium such as a magnetic disk or a CD-ROM, and is installed in the computer from these recording media or installed in the computer through a communication line.

  The acoustic echo canceling method and apparatus according to the present invention is utilized in the field of loudspeaker communication systems.

1 is a block diagram for explaining a first embodiment of the present invention. The flowchart for demonstrating the operation | movement procedure of Example 1 shown in FIG. The block diagram for demonstrating Example 2 of this invention. The block diagram for demonstrating Example 3 of this invention. The block diagram for demonstrating Example 4 of this invention. The block diagram for demonstrating the structure of the M channel echo cancellation part shown in FIG. FIG. 6 is a block diagram for explaining another configuration of the M channel echo cancellation unit shown in FIG. 5. The graph for demonstrating operation | movement of the nonlinearity determination part used as the principal part of this invention. The graph for demonstrating the other example of the graph shown in FIG. The block diagram for demonstrating the prior art.

Explanation of symbols

2,2 _{1 to} 2 _M speaker 41,41 _{1 to} 41 _N FIR filter
3,3 _{1 to} 3 _N microphone 42 Filter coefficient update means
4,4 ₁ to 4 _N Echo canceling part 43 1st subtraction means
5,5 _{1 to} 5 _N transmitter terminal 44,46 Non-linear processing means
6 DA converter 47 Second subtraction means
7,8 Amplifier 49 Adding means
9 AD converter

Claims (9)

A loudspeaker process for loudening the received signal and a sound pickup process for picking up the transmitted voice;
Filtering processing by an adaptive filter that generates a pseudo echo signal from the received signal;
A subtraction process for obtaining a difference signal from the pseudo echo signal from the collected sound signal and using it as a transmission signal;
A filter coefficient update process for updating a filter coefficient used for the filtering process by receiving the received signal and the difference signal;
When it is determined whether or not the nonlinear component is included in the collected sound signal, and when it is determined that the nonlinear component is not included, the filter coefficient update process is executed and it is determined that the nonlinear component is included , Nonlinearity determination processing for stopping the update operation of the filter coefficient update processing,
An acoustic echo canceling method comprising: A loudspeaker process for loudening the received signal and a sound pickup process for picking up the transmitted voice;
A received signal frequency band dividing process for dividing the received signal into subband received signals of a plurality of frequency bands;
A sound collection signal frequency band dividing process for dividing the sound collection signal into subband sound collection signals of a plurality of frequency bands;
The subband received signal and the subband collected signal divided by the received signal frequency band dividing process and the collected sound signal frequency band divided process are input, and a subband pseudo echo signal is generated from each subband received signal. Filtering processing by an adaptive filter, subtraction processing for obtaining a subband difference signal between a subband sound pickup signal and a subband pseudo echo signal, and an adaptive filter used for the above filtering processing with the subband received voice signal and the subband difference signal as inputs. Filter coefficient update processing for updating the coefficient, and whether or not the subband sound pickup signal includes a nonlinear component. If it is determined that the nonlinear component is not included, the filter coefficient update processing is executed, Nonlinearity determination that stops the filter coefficient update process when it is determined that a component is included And management executed for each subband, a subband echo canceling processing for outputting the subband difference signals from each of
A frequency band synthesis process for synthesizing a subband difference signal output by subband echo cancellation processing for each subband and making it a transmitted voice signal;
An acoustic echo canceling method comprising: A loudspeaker means for loudening the received signal and a sound pickup means for picking up the transmitted voice;
An adaptive filter that generates a pseudo echo signal from the received signal;
Subtracting means for obtaining a difference signal from the pseudo echo signal from the collected sound signal and using it as a transmission signal;
Filter coefficient updating means for updating a filter coefficient used in the adaptive filter with the received signal and the difference signal as inputs;
It is determined whether or not a linear component is included in the collected sound signal. When it is determined that a non-linear component is not included, an update operation of the filter coefficient updating unit is executed to include a non-linear component. A non-linearity determining means for stopping the updating operation of the filter coefficient updating means,
An acoustic echo canceling device comprising: The acoustic echo canceller according to claim 3,
A difference signal between a non-linearly processed sound pickup signal and a non-linearly processed pseudo echo signal is obtained and transmitted to a first subtracting means that obtains a difference signal from the sound pickup signal and a pseudo echo signal as a transmission signal. Second subtracting means for providing a speech signal is provided, and when the nonlinearity determining means determines that there is no nonlinear component in the collected sound signal, the difference signal output from the first subtracting means is used as a transmission signal, and the nonlinearity An acoustic echo canceling apparatus comprising: a switching unit that uses the difference signal output from the second subtracting unit as a transmission signal when it is determined by the determining unit that there is a nonlinear component. Either of the acoustic echo cancellers according to claim 3 or 4,
The nonlinearity determining means includes means for comparing the absolute value of the sound pickup signal with a preset threshold value and determining that a nonlinear component is included when the absolute value of the sound pickup signal is larger than the threshold value. Acoustic echo canceller. A loudspeaker means for loudening the received signal and a sound pickup means for picking up the transmitted voice;
A received signal frequency band dividing means for dividing the received signal into subband received signals of a plurality of frequency bands;
A sound collection signal frequency band dividing means for dividing the sound collection signal into subband sound collection signals of a plurality of frequency bands;
The subband received signal and the subband collected signal divided by each of the received signal frequency band dividing unit and the collected sound signal frequency band dividing unit are input, and a subband pseudo echo signal is generated from each subband received signal. A filter by an adaptive filter; a subtracting means for obtaining a subband difference signal between the subband sound pickup signal and the subband pseudo echo signal; and a filter coefficient used for the adaptive filter by inputting the subband received voice signal and the subband difference signal as input. Filter coefficient updating means for updating, and determining whether or not the non-linear component is included in the sub-band sound pickup signal. If it is determined that the nonlinear component is not included, the operation of the filter coefficient updating means is executed, A non-linearity determination process for stopping the operation of the filter coefficient updating means when it is determined that a component is included. Run for each sub-band, and sub-band echo canceling means for outputting the sub-band difference signals from each of
A frequency band synthesizing unit that synthesizes a subband difference signal output by the subband echo cancellation processing for each subband and makes a transmission voice signal;
An acoustic echo canceling device comprising: The acoustic echo canceller according to claim 6,
The subtracting means provided in the subband echo canceling section is a first subtracting means, and the subband of the subband pseudo echo signal subjected to nonlinear processing from the subband sound pickup signal subjected to nonlinear processing to the first subtracting means. Second subtracting means for obtaining a difference signal as a transmitted signal is provided, and the subband output from the first subtracting means when the non-linearity determining means determines that there is no non-linear component in the subband sound pickup signal. The difference signal is synthesized by the frequency band synthesizing unit to be a transmission signal, and the second subtracting unit outputs when the non-linearity judging unit judges that the subband sound pickup signal has a non-linear component. An acoustic echo canceling apparatus characterized in that a sub-band difference signal is synthesized by the frequency band synthesizing means into a transmission signal.   An acoustic echo canceling program written in a computer-readable program language and causing the computer to function as the acoustic echo canceling device according to any one of claims 3 to 7.   A recording medium constituted by a computer-readable recording medium and having the acoustic echo cancellation program recorded on the recording medium.

Images