JP5293342B2 - Voice communication apparatus, method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a voice communication device which does not degrade communication quality even if a residual echo suppression function overworks. <P>SOLUTION: The voice communication device includes an echo canceller which is composed of an echo canceller body that generates a quasi-echo to remove echo components included in transmission signals, and a residual echo component suppression means for suppressing residual echo components included in transmission signals that are output from the echo canceller body. In addition, the voice communication device includes a background noise component extraction means for extracting background noise components from transmission signals, and a noise superimposition means for superimposing the extracted background noise components on transmission signals that are output from the echo canceller. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a voice communication apparatus, method, and program, and uses, for example, an echo canceller used for canceling an echo generated between transmission and reception paths and a noise canceller used for canceling noise input from a transmission microphone. Applicable to the case.

  The telephone communication network has an adaptive echo canceller to remove echo (line echo) generated by a hybrid circuit (2-wire 4-wire conversion circuit) and echo (acoustic echo) that circulates from the speaker of the telephone terminal to the microphone. Widely used.

  In general, an echo canceller uses an adaptive filter to input a received signal from the other party's speaker and a transmitted signal mixed with an echo of the received signal, generate a pseudo echo, and subtract it from the transmitted signal. The method is taken. Furthermore, as part of the echo canceller function, there is also used a method of nonlinearly suppressing the level of echo that has remained unerased by subtraction processing using a pseudo echo estimated using an adaptive filter, using an attenuator, center clipper, etc. (See Patent Document 1).

  In addition to the echo canceller, there is also a voice communication apparatus having a noise canceller to remove noise picked up from the microphone.

  Furthermore, there is a voice communication apparatus having both functions of these echo canceller and noise canceller (see Patent Document 2). In such a voice communication apparatus having both functions, two patterns can be considered as insertion positions of the noise canceller. That is, there are a voice communication device that cancels the echo after canceling the noise first, and a voice communication device that cancels the noise after canceling the echo first.

  The echo canceller of the voice communication apparatus that cancels the echo after canceling the noise first receives the received signal from the far end input (Rin) end and the noise canceller output of the transmit signal from the near end input (Sin) end. To do.

  The echo canceller of the voice communication apparatus that cancels the noise after canceling the echo first receives the reception signal from the Rin end and the transmission signal from the Sin end. The output signal of the echo canceller is used as the input of the noise canceller.

JP-A-8-32494 JP 2002-118504 A

  However, when the echo canceller and the noise canceller are used together, they have the following problems.

  When canceling the echo after canceling the noise first, the filter for canceling the noise changes with time, and therefore, when learning the echo, there is a problem that accurate learning cannot be performed due to the influence of the noise canceller.

  On the other hand, when canceling the noise after canceling the echo first, there is a lot of error in the echo canceller output while the echo canceller is learning, and the noise canceller erroneously determines this as noise or speech, and determines the state and parameters. An error occurs in the estimation.

  In the voice communication device described in Patent Document 2, the effect of echo suppression and noise suppression is improved by passing parameters between the echo canceller and the noise canceller. However, preventing erroneous learning of the echo canceller and cancellation error of the noise canceller is incomplete, and the call quality may be deteriorated.

  The present invention has been made in view of the above points, and it is an object of the present invention to provide a voice communication device, method, and program that do not deteriorate the call quality even if the residual echo suppression function acts excessively. .

The first voice communication apparatus of this information includes (1) an echo canceller body that generates a pseudo echo and removes an echo component included in the transmission signal, and an output transmission signal from the echo canceller body. An echo canceller having residual echo component suppression means for suppressing the residual echo component in accordance with a nonlinear suppression characteristic ; and (2) extracting the background noise component by converting the transmission signal into a frequency domain signal. Background noise component extraction means for outputting the background noise component returned to the time domain signal, and (3) the background noise component extracted by the background noise component extraction means is added to the transmission signal output from the echo canceller. And a noise superimposing unit that superimposes the signal by a predetermined multiple of 1 and compensates the transmission signal in the excessive suppression period by the residual echo component suppressing unit .

In the voice communication method of the second aspect of the present invention, (1) the echo canceller body of the echo canceller generates a pseudo echo to remove the echo component contained in the transmission signal, and suppresses the residual echo component of the echo canceller. The means suppresses the residual echo component included in the output transmission signal from the echo canceller main body according to a nonlinear suppression characteristic , and (2) the background noise component extraction means converts the transmission signal into a frequency domain signal. A background noise component is extracted by conversion, and the extracted background noise component is converted back into a time domain signal for output . (3) The noise superimposing means adds the background noise component to the transmission signal output from the echo canceller. The background noise component extracted by the extraction means is superimposed by a first predetermined multiple, and the transmission signal in the excessive suppression period by the residual echo component suppression means is compensated .

The voice communication program according to the third aspect of the present invention provides a computer that (1) generates an echo echo and removes an echo component included in a transmission signal, and an output transmission from the echo canceller main body. An echo canceller having residual echo component suppression means for suppressing the residual echo component contained in the signal according to a nonlinear suppression characteristic ; and (2) extracting the background noise component by converting the transmission signal into a frequency domain signal. And a background noise component extracting means for outputting the extracted background noise component back to a time domain signal, and (3) background noise extracted by the background noise component extracting means to the transmission signal output from the echo canceller. superimposing the component first predetermined multiple to, that to function as the noise superimposing means for compensating a transmission signal suppression excessive period by the residual echo component suppressing means And butterflies.

  According to the present invention, since the background noise component can be superimposed in a period in which the residual echo suppression function is excessively applied, it is possible to prevent deterioration in call quality during that period.

It is a block diagram which shows the structure of the audio | voice communication apparatus which concerns on 1st Embodiment. It is a block diagram which shows the structural example of the noise extraction part in 1st Embodiment. It is a signal waveform diagram of each part before and after a single talk state by a far-end speaker under a situation where the level of background noise (noise) is large in the first embodiment. It is a block diagram which shows the structure of the audio | voice communication apparatus which concerns on 2nd Embodiment.

(A) First Embodiment Hereinafter, a first embodiment of a voice communication apparatus, method, and program according to the present invention will be described in detail with reference to the drawings.

(A-1) Configuration of First Embodiment FIG. 1 is a block diagram showing a configuration of a voice communication apparatus according to the first embodiment. For example, even if the voice communication device of the first embodiment is a soft phone and is realized by a CPU and a program executed by the CPU (voice communication program of the first embodiment), it is functionally Can be represented by the functional block diagram of FIG.

  In FIG. 1, a voice communication apparatus 1 according to the first embodiment is a voice communication apparatus having an echo canceller 4 for erasing acoustic echo that circulates from a speaker 2 of a telephone terminal to a microphone 3. Although the echo canceller 4 of the first embodiment shown in FIG. 1 is designed to cancel the acoustic echo, the echo canceller 4 of the first embodiment can also cancel the line echo generated in the hybrid circuit existing in the telephone communication network. good.

  The echo canceller 4 of the first embodiment includes an adaptive filter 101, a subtracter 102, and an attenuator 103.

  The adaptive filter 101 estimates an impulse response of an echo that goes from the speaker 2 to the microphone 3 and generates a pseudo echo.

  The subtracter 102 eliminates the echo component from the transmission signal transmitted by the voice communication apparatus 1 by subtracting the pseudo echo generated by the adaptive filter 101 from the transmission signal captured by the microphone 3.

  The attenuator 103 attenuates the level of the echo component that cannot be completely erased in the transmission signal output from the subtractor 102. The attenuator 103 receives the voice signal of the far-end speaker without receiving the voice signal of the far-end speaker from the near end to the microphone 3 and outputs the voice signal of the far-end talker to the speaker 2. It operates in the state of being in a single talk state of a far-end speaker.

  Although FIG. 1 shows the case where the attenuator 103 is applied as an attenuation of the level of the echo component that cannot be completely erased, the level of the echo component that cannot be completely erased is attenuated by another configuration. Anyway. For example, instead of the attenuator 103, a center clipper can be applied.

  Here, when the level of the background noise picked up by the microphone 3 is high, the attenuator 103 removes not only the remaining echo component but also the background noise. By removing such background noise, the other party's caller hears the voice being interrupted, creating a sense of incongruity in the call. When the background noise level picked up by the microphone 3 in a section where there is no voice signal of the far-end speaker and the near-end speaker is a level at which the suppression operation of the attenuator 103 does not work, the background noise is generated from the attenuator 103. Is output.

  In the case of the first embodiment, the processing for background noise can be selected by the near-end speaker. For example, the near-end speaker listens for hope from the far-end speaker and selects processing for background noise.

  In FIG. 1, the voice communication apparatus 1 according to the first embodiment has a noise extraction unit 104, a level variable unit 105, an adder 106, a level variable unit 107, a subtractor 108, and switches 109 and 110 as a processing configuration for background noise. Have Here, the noise extraction unit 104, the level variable unit 107, and the subtractor 108 constitute a noise canceller.

  A signal obtained by the microphone 3 is input to the noise extraction unit 104, and the noise extraction unit 104 generates (extracts) a noise (background noise) component based on the signal. As the noise extraction unit 104, a noise component extraction configuration in an existing noise canceller can be applied. FIG. 2 is a block diagram illustrating a configuration example of the noise extraction unit 104. In FIG. 2, an example of a method in the frequency domain using FFT (Fast Fourier Transform) is described, but the method is not limited to this method.

  In FIG. 2, the noise extraction unit 104 includes a window function processing unit 111, an FFT processing unit 112, a suppressor 113, an IFFT processing unit 114, and a window function overlap unit 115.

  The window function processing unit 111 inputs a signal acquired by the microphone 3 and extracts analysis interval data of FFT processing by window function processing.

  The FFT processing unit 112 receives the real time domain data output from the window function processing unit 111 and converts the time signal into frequency domain data of several FFT points by FFT processing.

  The suppressor 113 receives the frequency domain data output from the FFT processing unit 112, suppresses only the voice component without suppressing the noise component, and outputs the frequency domain data of only the noise component.

  The IFFT processing unit 114 is input with frequency domain data that includes only noise components by the suppressor 113, and inversely transforms the frequency domain data into real-time domain data by inverse FFT processing.

  The window function overlap unit 115 connects the data in the cut out state to continuous data and outputs it.

  The switches 109 and 110 are turned on alternatively and complementarily according to the operation of the user (near-end speaker).

  The switch 109 is turned on when, for example, it is desired to add noise to the signal portion nonlinearly attenuated by the attenuator 103 to eliminate the uncomfortable feeling of the call due to signal interruption. When the switch 109 is turned on, the level variable device 105 and the adder 106 function effectively.

  The level variable unit 105 varies the level of the noise component extracted by the noise extraction unit 104 when the switch 109 is turned on and supplies the level to the adder 106. The level variable device 105 changes the variable gain according to the operation of the user (near-end speaker).

  The adder 106 adds the noise component output from the level variable device 105 to the signal output from the attenuator 103. The adder 106 passes the signal output from the attenuator 103 as it is when the switch 110 is on.

  The switch 110 is turned on when, for example, the attenuation by the attenuator 103 is not performed (when the far-end speaker is not in a single talk state), the switch 110 is operated as a noise canceller that subtracts noise, and sound quality deterioration due to noise is to be eliminated. It is.

  The level variable unit 107 varies the level of the noise component extracted by the noise extraction unit 104 when the switch 110 is turned on, and supplies the level to the subtractor 108. The level variable unit 107 changes the variable gain according to the operation of the user (near-end speaker).

  The subtracter 108 subtracts the noise component output from the level variable unit 107 from the signal output from the attenuator 103. The subtractor 108 passes the signal output from the adder 106 as it is when the switch 109 is turned on.

(A-2) Operation of the First Embodiment Next, the operation of the voice communication device 1 according to the first embodiment will be described in the order of the operation when the switch 109 is on and the operation when the switch 110 is on. .

  In a situation where the level of background noise (noise) is large, attenuation by the attenuator 103 greatly suppresses the background noise, the signal to the far-end speaker is interrupted, and the far-end speaker feels uncomfortable with the call The user (near-end speaker) turns on the switch 109.

  At this time, the noise component extracted by the noise extraction unit 104 from the signal obtained by the microphone 3 is applied to the level variable device 105 via the switch 109, and the level is adjusted. The noise component adjusted in this way is added to the signal output from the attenuator 103 by the adder 105. By this addition, even if there is a break in the signal output from the attenuator 103, the break is eliminated, and the uncomfortable feeling of the call due to the break in the signal can be eliminated.

  FIG. 3 shows signal waveforms of respective parts before and after the single talk state of the far-end speaker in a situation where the level of background noise (noise) is large. As shown in FIG. 3A, the signal captured by the microphone 3 is obtained by superimposing an echo component and background noise (noise). Even after the echo removal operation using the pseudo echo, if the background noise level is large and the convergence of the filter coefficient of the adaptive filter 101 is poor, as shown in FIG. A signal superimposed with noise is output from the subtractor 102. Due to the non-linear suppression characteristic of the attenuator 103, the residual echo is largely suppressed in a section where the residual echo is large, and the output signal from the attenuator 103 does not include noise or an effective signal component in a certain section as shown in FIG. It becomes like this. By adding the noise extracted by the noise extraction unit 104 and level-adjusted by the level variable unit 105 to the output signal from the attenuator 103 by the adder 106, the voice communication apparatus 1 receives the signal shown in FIG. A signal including only noise whose level is substantially adjusted as shown in (D) is transmitted to the far end side.

  When the operation of the attenuator 103 does not cause the background noise (noise) to be interrupted, the user (near-end speaker) turns on the switch 110 so that the noise canceller functions effectively.

  In a state where neither the far-end speaker's audio signal nor the near-end speaker's audio signal exists, or in the state where only the far-end speaker's audio signal exists, the output signal from the attenuator 103 is almost background noise. It is meant to include only.

  When the switch 110 is turned on, the noise component extracted by the noise extraction unit 104 from the signal acquired by the microphone 3 is given to the level variable unit 107 via the switch 110, and the level is adjusted. The noise component adjusted in this way is subtracted from the signal output from the attenuator 103 by the subtractor 108, whereby the signal from which the noise has been removed is transmitted from the voice communication apparatus 1 to the far end side. Is done.

  Since the signal waveform diagram of each part when operating as a noise canceller is the same as that of a conventional noise canceller, its illustration is omitted.

(A-3) Effect of First Embodiment According to the first embodiment, even when an echo canceller having an auxiliary device that suppresses residual echo such as an attenuator or a center clipper is applied, large echo cancellation is performed. It is possible to make a call with little sense of interruption while securing the amount.

  Further, depending on the user's selection, it is possible to execute an operation weighted for noise removal instead of compensating for the discontinuity.

(B) Second Embodiment Next, a second embodiment of the voice communication apparatus, method and program according to the present invention will be described in detail with reference to the drawings.

  FIG. 4 is a block diagram showing the configuration of the voice communication apparatus according to the second embodiment, in which the same and corresponding parts as those in FIG. 1 according to the first embodiment are indicated by the same reference numerals. Yes.

  In the first embodiment, the switches 109 and 110 are manually operated by the user (near-end speaker). However, in the voice communication device 1A of the second embodiment, the switch control circuit 120 includes the switches 109 and 110. Controls the on / off state of.

  The switch control circuit 120, for example, (1) for a predetermined time from when the adaptive filter 101 starts an adaptive operation (or until it is considered to have converged), (2) in the single talk state where only the received signal exists, the attenuator 103 When the average level of the input level exceeds the predetermined threshold value (3) When the attenuation amount set in the attenuator 103 is large (for example, 6 dB), the switch 109 is turned on, and (1) to (3) Otherwise, the switch 110 is turned on.

  Except for such control of the switches 109 and 110, the configuration and operation of the second embodiment are the same as those of the first embodiment.

  According to the second embodiment, even when a device having an auxiliary device for suppressing a residual echo is applied, it is possible to make a call with a little sense of interruption while ensuring a large amount of echo cancellation. Instead of compensation, it is also possible to execute an operation weighted for noise removal.

  According to the second embodiment, there is an effect that such an operation mode can be automatically set.

(C) Other Embodiments In the description of each of the above-described embodiments, various modified embodiments have been referred to. However, modified embodiments as exemplified below can be given.

  In each of the above embodiments, the voice communication device having the noise canceller configuration is shown in addition to the configuration for compensating for the occurrence of the discontinuity due to the function of the auxiliary device for suppressing the residual echo, but the noise canceller configuration is not provided. A voice communication device may be configured.

  Further, in each of the above embodiments, the configuration for instantaneously switching between the configuration for compensating for the feeling of discontinuity and the configuration for the noise canceller has been shown, but it may be switched gradually. For example, the gain of one level variable device 105 or 107 that has been operating effectively is gradually reduced, and the gain of the other level variable device 107 or 105 that has been disabled is gradually increased. You may make it switch with.

  Further, in each of the above embodiments, the switches 109 and 110 are alternatively and complementarily turned on. However, the switches 109 and 110 may not be turned on at the same time. For example, as a modification of the second embodiment, instead of always turning on the switch 110 during the period when the switch 109 is not turned on, there is no received voice and the noise power is a predetermined threshold value during the period when the switch 109 is not turned on. The switch 110 may be turned on only at the above time.

  An apparatus to which the telephone communication apparatus of the above embodiment is applied is not limited to a telephone terminal, and may be a PBX or the like.

  DESCRIPTION OF SYMBOLS 1, 1A ... Voice communication apparatus, 4 ... Echo canceller, 104 ... Noise extraction part, 105, 107 ... Level variable device, 106 ... Adder, 108 ... Subtractor, 109, 110 ... Switch, 120 ... Switch control circuit.

Claims (6)

Echo canceller body that generates pseudo echo and removes echo component contained in transmission signal, and residual echo component contained in output transmission signal from echo canceller body is suppressed according to nonlinear suppression characteristics An echo canceller having residual echo component suppression means;
A background noise component extracting means for converting the transmission signal into a frequency domain signal to extract a background noise component, and returning the extracted background noise component to a time domain signal ;
The background noise component extracted by the background noise component extraction unit is superimposed on the transmission signal output from the echo canceller by a first predetermined multiple, and the transmission signal of the excessive suppression period by the residual echo component suppression unit is obtained. And a noise superimposing means for compensating . Noise removal means for removing the background noise component extracted by the background noise component extraction means by a second predetermined multiple from the transmission signal output from the echo canceller, and most of the noise superimposing means and the noise removal means. The voice communication apparatus according to claim 1, further comprising: a noise processing selection unit that makes only one of them effective. The echo canceller body of the echo canceller generates a pseudo echo to remove the echo component contained in the transmission signal, and the residual echo component suppression means of the echo canceller applies to the output transmission signal from the echo canceller body. Suppresses the residual echo component included according to the nonlinear suppression characteristics ,
The background noise component extraction means converts the transmission signal to a frequency domain signal to extract a background noise component, returns the extracted background noise component to a time domain signal, and outputs it,
The noise superimposing unit superimposes the background noise component extracted by the background noise component extracting unit on the transmission signal output from the echo canceller by a first predetermined multiple, and the excessive suppression period by the residual echo component suppressing unit A voice communication method characterized by compensating for the transmission signal . The noise removal means removes the background noise component extracted by the background noise component extraction means from the transmission signal output from the echo canceller by a second predetermined multiple,
4. The voice communication method according to claim 3, wherein the noise processing selecting means enables at least one of the noise superimposing means and the noise removing means at most. Computer
Echo canceller body that generates pseudo echo and removes echo component contained in transmission signal, and residual echo component contained in output transmission signal from echo canceller body is suppressed according to nonlinear suppression characteristics An echo canceller having residual echo component suppression means;
A background noise component extracting means for converting the transmission signal into a frequency domain signal to extract a background noise component, and returning the extracted background noise component to a time domain signal ;
The background noise component extracted by the background noise component extraction unit is superimposed on the transmission signal output from the echo canceller by a first predetermined multiple, and the transmission signal of the excessive suppression period by the residual echo component suppression unit is obtained. A voice communication program that functions as a noise superimposing means for compensation . Computer
Noise removal means for removing the background noise component extracted by the background noise component extraction means by a second predetermined multiple from the transmission signal output from the echo canceller, and most of the noise superimposing means and the noise removal means. 6. The voice communication program according to claim 5, wherein the voice communication program is made to function as a noise processing selection unit that enables only one of them.

Images