RU2799561C2 - Echo cancelling device, echo cancelling method and echo cancelling program - Google Patents

Abstract

FIELD: acoustics; devices for echo cancellation.

SUBSTANCE: echo canceller for suppressing an echo generated when a voice output from a speaker enters a microphone, comprising a level control unit provided in a received signal path for transmitting a received signal from a far end to a speaker; an echo removal unit provided in the transmission path for transmitting a signal input from the microphone, the echo removal unit removing residual echo from the captured audio signal output from the microphone; and a double talk detection unit that detects whether the signals are in a double talk state in which the signals are simultaneously transmitted to the transmitted signal path and the received signal path. The level adjusting unit includes a compressor that performs a compression process on a signal exceeding a first threshold among the received signals if the double talk detection unit determines a double talk state.

EFFECT: reduction of non-linear echoes, reduction of uncompensated echoes and improvement of the quality of conversational speech.

9 cl, 8 dwg

Description

Technical field

[0001]

The present invention relates to an echo canceling device, an echo canceling method, and an echo canceling program.

Prerequisites for the creation of the invention

[0002]

Patent Document 1 describes an echo canceling apparatus in which, when detecting no transmission of a signal through a transmission signal path and presence of a signal transmission through a received signal path, an echo canceller is used to perform an echo cancellation process on the captured audio signal.

List of bibliographic references

Patent Literature

[0003]

Patent Document 1: JP 2018-201147 A

Statement of the Invention

Technical task

[0004]

However, in the echo cancellation device described in Patent Document 1, if the performance of a speaker or speaker amplifier is low, there is a possibility of an increase in non-linear echoes, an increase in non-cancelled echoes, and a deterioration in speech quality.

[0005]

The present invention has been developed in view of the foregoing, and it is an object of the present invention to provide an echo cancellation apparatus, an echo cancellation method, and an echo cancellation program that can suppress non-linear echo and reduce degradation of speech quality.

Solution

[0006]

To solve this problem, the echo canceller according to the present invention is, for example, an echo canceller for suppressing an echo generated when a voice signal is output from a speaker to a microphone. The echo canceller includes a level adjusting unit, an echo canceling unit, and a double talk detection unit. A level control unit is provided in the received signal path for transmitting the received signal from the far end of the communication line to the speaker. An echo canceller is provided in the signal path for transmitting the input signal from the microphone. The echo remover removes the residual echo from the captured audio signal from the microphone. The double talk detection unit detects the presence of a double talk condition in which signals are transmitted simultaneously to the transmit signal path and to the receive signal path. The level control unit includes a compressor. If the double talk detection unit determines the double talk state, the compressor performs a compression process on a signal exceeding the first threshold among the received signals. If the double-talk detection unit does not determine the double-talk state, the compressor performs a compression process on a signal exceeding the second threshold, which exceeds the first threshold, among the received signals.

[0007]

The echo canceller according to the present invention includes a compressor in the received signal path for transmitting the received signal to the speaker. When determining the double talk state, the compressor performs a compression process on a signal that exceeds a first threshold among the signals received from the far end. Thus, non-linear echo can be prevented. As a result, deterioration in voice quality can be reduced.

[0008]

In this case, if the double-talk detection unit does not determine the double-talk state, the compressor may perform a despreading process on the signal exceeding the second threshold, which exceeds the first threshold, among the received signals. This makes it possible to more reliably prevent non-linear echo.

[0009]

Meanwhile, the level adjuster may include a gain adjuster that adjusts the gain of the received signal. The compressor may adjust the threshold such that the first threshold becomes small as the gain increases. The compressor may perform a compression process on the signal from the gain control unit. As a result, even if the gain adjuster outputs a high level voice signal, the compressor reduces the peak of the voice signal, and thus the non-linear echo can be reduced.

[0010]

In this case, the compressor can increase the compression ratio as the gain increases. As a result, even if the gain adjuster outputs a high level voice signal, the compressor reduces the peak of the voice signal, and thus the non-linear echo can be reduced.

[0011]

In this case, the compressor can change the compression ratio based on information about the distortion of the speaker. This allows you to reduce non-linear echo.

[0012]

In this case, an echo canceller can be further provided that performs an echo canceling process on a signal from which the residual echo has been removed by the echo canceller. As a result, even in a case where the sound level of the speaker is set large due to a large amount of noise in the environment or the like, and non-linear echoes are generated in large quantities, it is possible to remove an echo component.

[0013]

In this case, the compressor can compare the value of the received signal with the third threshold for each frequency band. If the double talk detection unit determines the double talk state, the compressor may perform a compression process on the received signal with a value greater than the third threshold. Accordingly, part of the signals compressed by the compressor is reduced to produce an even more natural voice, which improves the quality of a telephone conversation.

[0014]

To solve the problem, the echo cancellation method according to the present invention is, for example, an echo cancellation method for canceling echo in a near end terminal including a speaker and a microphone. The echo cancellation method includes: determining whether the signals are in a simultaneous talk state, in which the signals simultaneously pass into a transmitted signal path for transmitting a signal input from a microphone, and into a received signal path for transmitting a signal to a speaker; performing a despreading process on a signal exceeding a first threshold among the received signals from the far end when a double talk state is detected; outputting the signal after the compression process from the speaker; and removing the residual echo from the captured audio signal output from the microphone. As a result, it becomes possible to suppress the non-linear echo and eliminate the degradation of voice quality.

[0015]

To solve the problem, the echo cancellation program according to the present invention is, for example, an echo cancellation program for canceling echo in a near end terminal including a speaker and a microphone. The echo cancellation program causes the computer to function as: a double talk detection unit that detects whether signals are in a double talk state, in which signals are simultaneously transmitted to the transmit signal path for transmitting the signal input from the microphone, and to the signal path for transmitting the signal to the speaker ; a compressor that performs a compression process on a signal exceeding the first threshold among the received signals from the far end when determining the double talk state; and an echo removal unit that removes residual echo from the captured audio signal output from the microphone. As a result, the non-linear echo can be suppressed and deterioration of voice quality can be eliminated.

Benefits of the Invention

[0016]

According to the present invention, the non-linear echo can be suppressed and voice degradation can be eliminated.

Brief description of graphic materials

[0017]

FIG. 1 is a diagram roughly showing a voice communication system 100 with an echo canceller 1 according to the first embodiment.

FIG. 2 is a block diagram showing the general configuration of the echo canceller 1.

FIG. 3 is a diagram schematically illustrating a process carried out by a compressor when a double talk state is detected.

FIG. 4 is a diagram schematically illustrating a process carried out by a compressor when a double talk state is not detected.

FIG. 5 is a diagram schematically illustrating the process carried out by the compressor when the gain is changed.

FIG. 6 is a block diagram illustrating the general configuration of the echo canceller 2.

FIG. 7 is a block diagram showing the general configuration of the echo canceller 3.

FIG. 8 is a diagram schematically illustrating the process carried out by the compressor.

Description of Embodiments

[0018]

Embodiments of the echo canceller according to the present invention will be described in detail below with reference to the drawings. An echo canceller is a device that suppresses the echo generated during a telephone call in a voice communication system.

[0019]

First Embodiment

FIG. 1 is a diagram roughly showing a voice communication system 100 with an echo canceller 1 according to the first embodiment. The voice communication system 100 mainly includes a terminal 50 including a microphone 51 and a speaker 52, two cell phones 53 and 54, a speaker amplifier 55, and an echo canceller 1.

[0020]

The voice communication system 100 is a system in which a user (user A on the near end side) using a terminal 50 (near end terminal) is in voice communication with a user (user B on the far end side) using a cell phone 54 (terminal far end). The input voice signal by the cell phone 54 is amplified and output by the speaker 52, where the microphone 51 receives the voice signal given by the user on the near end side and transmits to the cell phone 54. Thus, user A can make an amplified voice call (hands-free call) without holding the cell phone 53 in hand. Cell phone 53 and cell phone 54 are connected to each other via a public telephone line.

[0021]

The echo canceller 1 suppresses echoes generated when the voice signals output from the speaker 52 to the microphone 51. The echo canceller 1 is provided between the terminal 50 and the cellular phone 53, i. in the transmit signal path for transmitting the audio signal picked up by the microphone 51 from the microphone 51 to the cell phone 53 and in the received signal path for transmitting the received signal from the cell phone 54 on the far end side from the cell phone 53 to the speaker 52.

[0022]

The echo cancellation device 1 can be made in the form of a specialized board installed on a speech terminal or the like. (eg, an embedded device, a conferencing system, or a mobile terminal) in the voice communication system 100. Further, the echo canceller 1 may include, for example, computer hardware and software (an echo canceller). The echo canceller may be stored in advance, for example, in an HDD as a storage medium built into a device such as a computer, and in a ROM in a microcomputer containing a central processing unit (CPU), and may be installed thereon to the computer. In addition, the echo canceller may be temporarily or permanently stored (memorized) in a removable storage medium such as a semiconductor memory device, a memory card, an optical disk, a magneto-optical disk, a magnetic disk, or the like.

[0023]

FIG. 2 is a block diagram showing the general configuration of the echo canceller 1. The echo canceller 1 mainly includes a level adjustment section 11 , an echo cancellation section 13 , and a double talk detection section 15 . FIG. 2, the upper signal path is a transmitted signal path and the lower signal path is a received signal path.

[0024]

Block 11 level adjustment is provided in the path of the received signal. The level control unit 11 mainly includes a gain control 111 and a compressor 112.

[0025]

The gain controller 111 is a gain adjuster that adjusts the gain of an input received signal. Specifically, the gain controller 111 adjusts the gain level (gain) of the input signal to adjust the level (amplitude) of the output signal. The gain controller 111 may automatically change the gain depending on the noise or the like. in the environment in which the terminal 50 is installed. Further, by operating an input unit such as a control knob, the gain controller 111 may change the gain depending on the position of the input unit.

[0026]

The output signal from the gain controller 111 is provided to the compressor 112. The compressor 112 amplifies (i.e., compresses) the over-threshold received signal from among the input received signals by a predetermined factor (the factor has a value less than 1) and outputs the signal. Compressor 112 will be described in detail below.

[0027]

It should be noted that in the present embodiment, the level control section 11 includes a gain control 111 and a compressor 112, but the gain control 111 is optional. In the absence of the gain control 111, the received signal transmitted from the cell phone 53 goes directly to the compressor 112, and the compressor 112 only needs to compress the received signal above the threshold from among the input received signals.

[0028]

An echo canceller 13 is provided in the transmission path to remove residual echo from the captured audio signal output from the microphone 51. The echo canceller 13 is a linear echo canceller that removes residual echo with an adaptive filter. More specifically, the echo canceling unit 13 updates the filter coefficient according to a predetermined procedure to generate a pseudo-echo signal from a signal transmitted through the received signal path, and subtracts the pseudo-echo signal from the signal transmitted through the transmitted signal path to remove residual echo. It should be noted that adaptive filters are well known, and thus description of the adaptive filter is omitted.

[0029]

It should be noted that in the present embodiment, the adaptive filter is applied to the echo canceller 13, but another known echo canceling algorithm can be applied to the echo canceller 13.

[0030]

After the residual echo is removed from the signal by the echo removal unit 13, the signal is transmitted to the cell phone 53. The signal from which the residual echo has been removed by the echo removal unit 13 is an input signal to the double talk detection unit 15 .

[0031]

The double talk detection unit 15 determines whether the input voice signal transmitted to the echo canceller 1 is in the single talk state or the double talk state. In this case, the term "single talk" refers to a state (near-end speech or far-end speech) in which either user A or user B emits a voice, and the signal is transmitted either to the transmitted signal path or to the received signal path. signal. Simulcast refers to a state (near-end speech or far-end speech) in which both user A and user B emit voices, and signals are simultaneously transmitted to the transmit signal path and the receive signal path.

[0032]

For example, the double talk detection unit 15 stores a frequency mask generated based on the training signal. The training signal is the signal transmitted through the transmission path during one-way speech (single talk) at the far end, where microphone 51 receives only the sound output from speaker 52. training signals.

[0033]

The double talk detection unit 15 compares the power spectrum value of the captured audio signal with the frequency mask value for each frequency band. When the number of frequency bands in which the value of the captured audio signal exceeds the value of the frequency mask is equal to or greater than a constant value, the sound input from the microphone 51 is detected and the signal is transmitted (there is speech at the near end) through the transmission signal path. The double talk detection unit 15 compares the power spectrum value of the received signal with the frequency mask value for each frequency band. When the number of frequency bands in which the received signal value exceeds the frequency mask value is equal to or greater than a constant value, signal transmission is detected (there is speech at the far end) through the received signal path.

[0034]

However, the double talk detecting unit 15 can determine whether the signal is in the single talk state or in the double talk state using various other known methods.

[0035]

Compressor 112 will be described in detail below. The results are fed to the compressor 112 from the double talk detection unit 15 . Compressor 112 performs different processes depending on whether the signal is in the double talk state.

[0036]

FIG. 3 is a diagram schematically illustrating the process performed by the compressor 112 in determining the double talk state. If the double talk detection unit 15 detects the double talk state, the compressor 112 performs a compression process on a signal exceeding the threshold I from among the received signals.

[0037]

FIG. 4 is a diagram schematically illustrating the process performed by the compressor 112 when a double talk condition is not detected. If the double-talk detection unit 15 does not detect the double-talk state, the compressor 112 performs a despreading process on a signal exceeding the threshold II among the received signals. Threshold II is greater than threshold I.

[0038]

When there is a double talk state, the operation performed by the echo cancellation unit 13 is generally unstable. Thus, the threshold I is reduced and the peak of the voice signal from the speaker 52 is reduced so that the echo cancellation unit 13 is reliably operated. In contrast, in the absence of a double-talk condition, since a sensation of discomfort is likely to be felt in the voice due to a too low peak of the voice signal, a Threshold II greater than Threshold I is used to maintain voice quality.

[0039]

FIG. 3 and 4, the solid line shows the signal before the compression process by the compressor 112, and the dashed line shows the signal after the compression process by the compressor 112. In the compression process, the compressor 112 multiplies the received signal exceeding threshold I or threshold II among the received signals inputted from the gain controller 111 by a factor of 1 or less which is set to the received signal, thus the output level can be reduced.

[0040]

As a result, it is possible to reduce the distorted sound generated by strong vibration of the speaker 52, the case holding the speaker 52, a component provided in the case, or the like. In particular, if the efficiency of the speaker 52 or the amplifier of the speaker 55 is low, the terminal 50 is small, or the like, a distorted sound is likely to be generated due to vibrations of the speaker 52 or the like, but lowering the voice level greatly reduces the distorted sound.

[0041]

By reducing the distorted sound on the speaker 52, the signals generated by picking up sounds by the microphone 51 and feeding into the echo canceller 13 are unlikely to contain non-linear echoes, and the echo canceller 13 can cancel echoes sufficiently.

[0042]

In addition, the compressor 112 does not perform a compression process on a received signal that is less than threshold I or threshold II, and outputs the input signal as is. As a result, the unpleasant feeling due to the volume change of the speaker 52 and the interruption of the voice is greatly reduced.

[0043]

According to the present embodiment, the compression process for a signal exceeding threshold I or threshold II can prevent non-linear echoes and stabilize the operation of the echo canceller 13 . This makes it possible to reduce degradation of voice quality while reducing uncompensated echoes.

[0044]

In addition, according to the present embodiment, a non-linear echo is less likely to occur, so it is only necessary to provide an echo removal unit 13 that removes a linear echo, and the amount of calculation required to remove echoes can be reduced.

[0045]

For example, in the absence of compressor 112, if the non-linear echo is large and the echo needs to be cancelled, an echo canceller is needed, as in the prior art. Thus, a large number of calculations are always required, which slows down the implementation of processes. For example, it is also understandable to use an echo canceller using a non-linear adaptive filter such as a Volterra filter to suppress non-linear echoes. However, a huge number of calculations are required (10 or more times more than in the case of a linear echo canceller).

[0046]

On the contrary, according to the present embodiment, by reducing the vibration of the speaker 52 or the like. the difference between the received signal received by the receiving side and the voice signal output from the speaker 52 is reduced, thus non-linear echoes can be reduced and echoes can be sufficiently removed by only the echo canceller 13, which is a linear echo canceller.

[0047]

It should be noted that in the present embodiment, when the double talk state is determined, the compressor 112 performs a despreading process on the received signal greater than the I threshold, and when no double talk state is detected, the despreading process is performed on the received signal greater than the threshold II (which is greater than the I threshold). However, if the double talk state is not detected, it is not necessary to perform a despreading process on a received signal that exceeds threshold II. However, in order to prevent degradation of voice quality when non-linear echoes are reduced, it is desirable to perform the compression process in both cases, when a double talk state is detected and when a double talk state is not detected.

[0048]

In addition, in the present embodiment, if the double talk detection unit 15 detects the double talk state, the compressor 112 performs a compression process on a signal exceeding the threshold I among the received signals, but the threshold applied in the double talk state may vary depending on situations.

[0049]

For example, when determining the variable threshold as the threshold I, the compressor 112 sets the threshold Ia when the gain for amplifying the received signal in the gain controller 111 is a certain value (set as the value a) as the threshold Ia. When the gain becomes larger than the value a, the threshold I is set less than the threshold Ia, and when the gain becomes smaller than the value a, the threshold I is set greater than the threshold Ia. However, the maximum value of threshold I is set to be less than threshold II. As a result, even if a large voice signal is output from the gain control 111, non-linear echoes can be reduced by reducing the peak of the voice signal by the compressor 112.

[0050]

In addition, in the present embodiment, although the ratio (value less than 1) used by the compressor 112 in the compression process is constant, the ratio used in the compression process may vary depending on the situation.

[0051]

FIG. 5 is a diagram schematically illustrating the process carried out by the compressor 112 when changing the gain. For example, when determining the variable gain as the gain b, the compressor 112 sets the gain b when the gain for amplifying the received signal in the gain controller 111 is the value a as the gain c. As the gain becomes larger than the value a, the factor b is set less than the factor c, and as the gain becomes smaller than the value a, the factor b is set greater than the factor c. In this case, if the reduction percentage in the output signal level during compression is defined as the compression ratio, the compression ratio increases as the ratio decreases. Thus, as the gain increases, the compression ratio increases, and as the gain decreases, the compression ratio decreases. As a result, even if a large voice signal is output from the gain control 111, non-linear echoes can be reduced by reducing the peak of the voice signal by the compressor 112.

[0052]

For example, the compressor 112 may change the compression ratio based on the distortion information of the speaker 52. In this case, the distortion information of the speaker 52 is, for example, a total harmonic distortion (or harmonic distortion) representing the degree of signal distortion. The total harmonic distortion factor at a low value indicates a small distortion of the speaker 52 and a large distortion of the speaker 52 at a large value of the factor. Thus, the compressor 112 can increase the compression ratio when the total harmonic distortion is high, and decrease the compression ratio when the total harmonic distortion is reduced. As a result, when using speaker 52, which may experience distortion, reducing the peak of the voice signal by compressor 112 can reduce non-linear echoes.

[0053]

Second Embodiment

The second embodiment of the present invention has a configuration in which an echo canceller is provided. Next, the echo canceller 2 according to the second embodiment will be described. The echo canceller 2 is particularly suitable for the case where the environment of a built-in device or the like is can change greatly. It should be noted that the same components as those in the echo canceller 1 according to the first embodiment are assigned the same reference numbers, and description of these components will be omitted.

[0054]

FIG. 6 is a block diagram showing the general configuration of the echo canceller 2. The echo canceller 2 mainly comprises a level adjusting unit 11, an equalizer 12, an echo canceling unit 13, an echo canceller 14, a double talk detection unit 15, a noise estimation unit 16, a noise suppression unit 17, and an equalizer 18.

[0055]

Equalizers 12, 18 raise or lower a particular frequency band of the voice signal. However, equalizers 12, 18 are optional.

[0056]

The echo canceller 14 performs a fast Fourier transform on the signal after the linear echo is removed by the echo canceller 13, performs an echo cancellation process (strong echo cancellation process) on the signal after performing the fast Fourier transform, and performs an inverse fast Fourier transform on the signal after performing the echo cancellation process to remove the non-linear echo. The processing for echo cancellation is well known, and thus a detailed description of the processing is omitted.

[0057]

If the double talk detection unit 15 determines that there is no signal transmission through the transmitted signal path, but a signal is transmitted through the received signal path, the echo canceller 14 may perform an echo cancellation process for the signal from which the residual echo has been removed by the echo removal unit 13.

[0058]

However, in the present embodiment, since the compressor 112 reduces non-linear echoes and the echo canceller 13 sufficiently removes the echo components, the echo canceller 14 actually works only when the sound level of the speaker 52 is set to high, for example, in the case where the external noise voice level is large and generates a lot of non-linear echoes.

[0059]

It should be noted that in the present embodiment, the echo canceller 14 performs frequency analysis using the Fast Fourier Transform, but instead of the Fast Fourier Transform, the Discrete Fourier Transform (DFT) can be used for frequency analysis. In addition, the echo canceller 14 may perform an inverse discrete Fourier transform instead of a fast Fourier transform.

[0060]

The noise estimator 16 estimates the noise component contained in the remote echo signal converted into a function of the frequency domain by the echo canceller 14, namely the estimated noise signal, for each frequency domain, and estimates the signal-to-noise ratio (SN) of the remote echo signal based on the power spectral density of the estimated noise signal that has been estimated. The noise suppressor 17 suppresses the noise signal in the remote echo signal based on the power spectral density of the estimated noise signal estimated by the noise estimator 16, and generates a noise canceled signal. It should be noted that the noise estimation unit 16 and the noise suppression unit 17 are optional.

[0061]

According to the present embodiment, it is possible to remove an echo component even when a plurality of non-linear echoes are generated. For example, in an embedded device, the environment may change greatly when driving a vehicle. When there is a lot of external noise, listening to the sound coming from the speaker 52 is difficult. Therefore, it is necessary to increase the volume of the sound of the speaker 52 and increase the volume of the voice of the user B on the far end side, resulting in an increase in linear echoes and non-linear echoes. The echo canceller 13 can remove linear echoes, but cannot remove non-linear echoes. In the present embodiment, the echo canceller 14 makes it possible to remove echo components even when a plurality of non-linear echoes are generated.

[0062]

Third Embodiment

The third embodiment has a configuration in which compressor operation differs depending on each frequency band. Next, the echo canceller 3 according to the third embodiment will be described. It should be noted that the same components as those in the echo canceller 1 according to the first embodiment are assigned the same reference numbers, and description of these components will be omitted.

[0063]

FIG. 7 is a block diagram showing the general configuration of the echo canceller 3. The echo canceller 3 mainly includes a level adjusting section 11, an echo canceling section 13, and a double talk detection section 15. The level control unit 11A mainly includes a gain control 111 and a compressor 112A.

[0064]

Compressor 112A compares the voice level with a threshold for each frequency band and performs a compression process on the received signal in the frequency band in which the voice level exceeds the threshold. Compressor 112A includes a processing unit that performs a Fourier transform and an inverse Fourier transform.

[0065]

The compressor 112A Fourier transforms the received signal to divide the power, which is the average energy per unit time, into the power for each frequency band, and calculates a power spectrum that expresses the power for each frequency band versus frequency for each unit time interval. Compressor 112A compares the value of the received signal with a threshold for each frequency band, and performs a despreading process on a signal that is greater than or equal to the threshold in the frequency band.

[0066]

FIG. 8 is a diagram schematically illustrating the process carried out by the compressor 112A. The solid line in FIG. 8 indicates the received signal. If the double talk detection unit 15 detects the double talk state, the compressor 112A performs a despreading process for a signal in a frequency band in which the signal value exceeds the threshold III. The dotted line in FIG. 8 indicates a signal after compressor 112A has performed a compression process on a signal exceeding threshold III. Moreover, if the double-talk detection unit 15 does not detect the double-talk state, the compressor 112A performs a despreading process for a signal in a frequency band in which the signal value exceeds the threshold IV. The dotted line in FIG. 8 indicates a signal after compressor 112A has performed a compression process on a signal exceeding threshold IV. Threshold IV is greater than threshold III.

[0067]

Compressor 112A outputs a signal obtained by performing an inverse Fourier transform on the signal after compression.

[0068]

According to the present embodiment, by changing the presence or absence of the compression process for each frequency band, it is possible to reduce the proportion of signals compressed by the compressor 112 to obtain an even more natural voice, thus improving the quality of a telephone conversation.

[0069]

It should be noted that in the present embodiment, although the compressor 112A performs a band-wide compression process on a signal that exceeds threshold III or threshold IV regardless of the frequency band, the threshold can also be changed depending on the frequency band. For example, a voice component at a low frequency is likely to cause distortion, and so the threshold can be reduced as the frequency decreases, and as the frequency increases, the threshold can be increased.

[0070]

The embodiments of the invention have been described in detail above with reference to the drawings. However, the specific configurations are not limited to the embodiments shown, but also include changes in design or the like without departing from the gist of the invention.

List of symbols

[0071]

1, 2, 3 - Echo Canceller

11, 11A - Level control unit

12 - Equalizer

13 - Echo remover

14 - Echo Canceller

15 - Double talk detection block

16 - Noise estimation block

17 - Noise suppression block

18 - Equalizer

50 - Terminal

51 - Microphone

52 - Speaker

53 - Cell phone

54 - Cell phone

55 - Speaker amplifier

100 - Voice communication system

111 - Gain control

112, 112A - Compressor.

Claims (27)

1. An echo cancellation device for suppressing the echo generated when an output voice signal arrives from a speaker into a microphone, comprising: a level control unit provided in the received signal path for transmitting the received signal from the far end to the speaker; an echo removal unit provided in the transmission path for transmitting a signal input from the microphone, the echo removal unit removing residual echo from the captured audio signal output from the microphone; And a double talk detection unit that determines whether the signals are in a double talk state, in which the signals are simultaneously transmitted to the transmitted signal path and to the received signal path, wherein the level adjusting unit includes a compressor that performs a compression process on a signal exceeding a first threshold among the received signals if the double talk detection unit detects a double talk state. 2. The echo cancellation device according to claim 1, wherein if the double talk detection unit does not determine the double talk state, the compressor performs a compression process on the signal exceeding the second threshold, which exceeds the first threshold, among the received signals. 3. An echo cancellation device according to claim 1 or 2, wherein the level adjuster includes a gain adjuster that adjusts the gain of the received signal, and the compressor adjusts the threshold such that the first threshold becomes small as the gain increases, and the compressor performs a compression process on the output signal from the gain adjuster. 4. The echo cancellation device according to claim 3, wherein the compressor increases the compression ratio as the gain increases. 5. The echo cancellation device according to any one of paragraphs. 1-3, in which the compressor changes the compression ratio based on the speaker distortion information. 6. The echo cancellation device according to any one of paragraphs. 1-5, further comprising: an echo canceller that performs a process for canceling an echo in a signal from which the residual echo has been removed by the echo canceller. 7. The echo cancellation device according to any one of paragraphs. 1-6, in which the compressor compares the received signal value with the third threshold for each frequency band, and if the double talk detection unit determines the double talk state, the compressor performs a compression process for the received signal with a value greater than the third threshold. 8. An echo cancellation method for echo cancellation in a near end terminal containing a speaker and a microphone, comprising: determining whether the signals are in a double talk state, wherein the signals are simultaneously transmitted to a transmit signal path for transmitting an input signal from a microphone and to a received signal path for transmitting a signal to a speaker; performing a compression process on a signal exceeding the first threshold among the received signals from the far end when determining the double talk state; outputting the signal after the compression process from the speaker; And removing residual echo from the captured sound output from the microphone. 9. Removable media that stores echo cancellation software for echo suppression in a near-end terminal containing a speaker and microphone, the echo cancellation software causing the computer to function as: a double talk detection unit that detects whether the signals are in a double talk state in which signals are simultaneously transmitted to a transmission signal path for transmitting a signal input from a microphone and to a received signal path for transmitting a signal to a speaker; a compressor that performs a compression process on a signal exceeding the first threshold among the received signals from the far end when determining the double talk state; And an echo removal unit that removes the residual echo from the captured audio signal output from the microphone.

Images