JP2010268378A - Echo canceller and acoustic echo eliminating method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To suppress voice distortion and enhance acoustic echo suppression effect. <P>SOLUTION: In an echo canceller 1, an output synthesizing section 109 adjusts acoustic pressure levels of output LOut (t, f) of a linear echo canceller section 106 that suppresses acoustic echo of a near-end voice signal and output NLOut (t, f) of a nonlinear echo canceller section 107 that suppresses residual echo which is the acoustic echo remaining in the near-end voice signal output from the linear echo canceller section 106, depending on acoustic echo suppression effect E<SB>linear</SB>(t) of the linear echo canceller section 106, which is measured by an acoustic-echo suppression effect measurement section 108, and generates a final output Out (t, f) by synthesizing both outputs. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an acoustic echo cancellation technique, and more particularly to an acoustic echo cancellation technique suitable for a conference system having a microphone and a speaker.

  In Patent Document 1, in a telephone conference system or a video conference system, a microphone array having a plurality of microphone elements is used to determine a band in which only speaker sound exists from a phase difference between microphones, and an acoustic echo canceller is applied only to that band. A technique for adaptive control is disclosed. According to this technique, the acoustic echo canceller can be dynamically controlled according to the situation of the conference room or the like, and the acoustic echo suppression effect can be enhanced.

JP 2008-141718 A

  The acoustic echo canceller described in Patent Document 1 is a linear echo canceller that performs processing for suppressing an acoustic echo component having a linear characteristic with respect to frequency, generates a pseudo echo signal from an audio signal, and uses the pseudo echo signal as an audio signal. The acoustic echo is suppressed by subtracting from. Here, the pseudo echo signal is generated from the audio signal based on the estimation result of the acoustic path. However, it is difficult to perform a complete estimation in an actual usage environment. For this reason, an estimation error actually occurs and an acoustic echo remains.

  For this reason, in the technique described in Patent Document 1, the size of the acoustic echo occupied in the input audio signal is estimated for each frame, and transmission of frames in which this is equal to or greater than a predetermined threshold is blocked by a voice switch (patent In this way, although it is instantaneous, the transmission of the voice signal is interrupted and the full-duplex call state cannot be maintained.

  Therefore, a nonlinear echo canceller (for example, spectral subtraction) that suppresses acoustic echo components with nonlinear characteristics with respect to frequency is provided to suppress residual echo, which is an acoustic echo remaining in the audio signal output from the linear echo canceller. A way to do it is considered. However, since the nonlinear echo canceller changes the sound pressure level of the sound signal output from the linear echo canceller for each frequency band, the sound distortion increases as the suppression amount increases.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide an acoustic echo cancellation technique capable of suppressing sound distortion and enhancing the acoustic echo suppression effect.

  In order to solve the above problems, the present invention provides a linear echo canceller that suppresses acoustic echo of a speech signal, a non-linear echo canceller that suppresses residual echo that is an acoustic echo remaining in the speech signal output from the linear echo canceller, and And adjusting the sound pressure levels of the outputs of the linear echo canceller and the nonlinear echo canceller according to the echo suppression effect of the linear echo canceller, and synthesizing them.

For example, the present invention is an echo canceller that eliminates an acoustic echo of an output audio signal that is an audio signal output from a speaker from an input audio signal that is an audio signal input to a microphone.
Pseudo echo generation means for estimating the acoustic echo of the output audio signal and generating a pseudo echo signal;
Linear echo cancellation means for subtracting the pseudo echo signal from the input audio signal to suppress acoustic echo of the output audio signal contained in the input audio signal;
A residual echo that is a residual component of the acoustic echo included in the input speech signal by changing a sound pressure level for each frequency band for the input speech signal in which the acoustic echo is suppressed by the linear echo canceling unit. A nonlinear echo canceling means for suppressing
Suppression effect measuring means for measuring the suppression effect of the acoustic echo by the linear echo cancellation means;
Based on the measurement result of the suppression effect measuring unit, the input audio signal in which the acoustic echo is suppressed by the linear echo canceling unit and the input audio signal in which the residual echo is suppressed by the nonlinear echo canceling unit Output synthesizing means for varying the sound pressure level and synthesizing both.

  According to the present invention, sound distortion can be suppressed and the acoustic echo suppression effect can be enhanced.

FIG. 1 is a schematic configuration diagram of an echo canceller according to an embodiment of the present invention. FIG. 2 is a diagram illustrating an example of a graph representing the relationship between Y (t) and E _linear (t) when Y (t) = 20 log (X (t)).

  The embodiment of the present invention will be described by taking as an example the case where the present invention is applied to an echo canceller of a microphone array device used in a conference system having a microphone and a speaker. Note that the microphone array device has a function of transmitting and receiving audio signals via an IP network and an acoustic echo canceller function.

  FIG. 1 is a schematic configuration diagram of an echo canceller according to an embodiment of the present invention.

  As illustrated, the echo canceller 1 includes a far-end speech signal input unit 101, a near-end speech signal input unit 102, FFT units 103 and 104, a pseudo echo generation unit 105, a linear echo canceller unit 106, a nonlinear An echo canceller unit 107, an acoustic echo suppression effect measurement unit 108, an output synthesis unit 109, an IFFT unit 110, and a near-end audio signal output unit 111 are included.

  The far-end audio signal input unit 101 is an input terminal for a digital audio signal (hereinafter referred to as a far-end audio signal) received from a call partner. Note that the far-end audio signal is output from the speaker after DA conversion.

  The near-end audio signal input unit 102 is an input terminal for a digital audio signal (hereinafter referred to as a near-end audio signal) input to the microphone and subjected to AD conversion. The near-end voice signal is transmitted to the other party after the acoustic echo is erased by the echo canceller 1.

  The FFT unit 103 performs FFT (Fast Fourier Transform) on the far-end audio signal input to the far-end audio signal input unit 101 and outputs frequency domain information of the far-end audio signal.

  The FFT unit 104 performs FFT on the near-end audio signal input to the near-end audio signal input unit 102 and outputs frequency domain information of the near-end audio signal.

  The pseudo echo generation unit 105 estimates the acoustic echo component of the far-end speech signal for each predetermined frequency band using an adaptive filter based on the frequency domain information of the far-end speech signal output from the FFT unit 103, Generate frequency domain information of the pseudo echo signal.

  When the reference signal (here, the far-end audio signal) is input to the filter, the adaptive filter is a change that occurs until the reference signal is output from the speaker and transmitted as a sound wave to reach the microphone. As long as it is a filter created to output the result (signal) of applying an equivalent change to its reference signal. Such an adaptive filter is created as a result of acoustic environment learning performed in accordance with the call situation. Various existing techniques including the technique described in Patent Document 1 can be used for the pseudo echo generation unit 105 and the adaptive filter.

  The linear echo canceller unit 106 subtracts the frequency domain information of the pseudo echo signal from the frequency domain information of the near end audio signal output from the FFT unit 104, thereby suppressing the acoustic echo included in the near end audio signal.

  The non-linear echo canceller unit 107 calculates a sound pressure level for each predetermined frequency band with respect to the sound pressure spectrum information of the near-end speech signal whose acoustic echo is suppressed by the linear echo canceller unit 106, and the sound pressure of the pseudo echo signal. It suppresses according to the sound pressure level in the same frequency band of spectrum information. The nonlinear echo canceller unit 107 can use an existing spectral subtraction technique.

  The acoustic echo suppression effect measurement unit 108 measures the acoustic echo suppression effect of the linear echo canceller unit 106 based on the ratio of the output of the linear echo canceller unit 106 to the input of the linear echo canceller unit 106. Details of the acoustic echo suppression effect measuring unit 108 will be described later.

  Based on the acoustic echo suppression effect measured by the acoustic echo suppression effect measurement unit 108, the output synthesis unit 109 changes the sound pressure level of each of the output of the linear echo canceller unit 106 and the output of the nonlinear echo canceller unit 107, and Synthesize. Details of the output combining unit 109 will be described later.

  The IFFT unit 110 performs IFFT (Inverse FFT) on the frequency domain information of the near-end audio signal output from the output synthesis unit 109, and outputs a near-end audio signal.

  The near-end audio signal output unit 111 is an output terminal for the near-end audio signal output from the IFFT unit 110.

  Next, details of the acoustic echo suppression effect measuring unit 108 will be described.

In the following description, the input to the linear echo canceller unit 106 (near-end speech signal in the frequency band f at the frame time t) is In (t, f) and In (t, f). The output (near-end speech signal after acoustic echo suppression) is LOut (t, f). Further, the pseudo echo signal generated by the pseudo echo generation unit 105 from the far-end voice signal in the frequency band f at the frame time t is used as the estimated echo of the linear echo canceller unit 106 for Echo (t, f) and In (t, f). The suppression amount is P _estm (t, f), and the acoustic echo suppression effect of the linear echo canceller unit 106 for the near-end speech signal at frame time t is E _linear (t) [dB].

First, the acoustic echo suppression effect measuring unit 108 calculates P _estm (t, f) for each frequency band (f = 0 to n−1, where n is the number of frequency band divisions) for each frame as follows.

  The acoustic echo suppression effect measurement unit 108 compares In (t, f) and Echo (t, f) to determine whether the acoustic echo is dominant in In (t, f). As the proportion of acoustic echo in In (t, f) increases, Echo (t, f) approaches In (t, f). Therefore, whether or not the acoustic echo is dominant is determined based on whether or not the ratio of Echo (t, f) to In (t, f) is equal to or greater than a predetermined threshold. Specifically, it is possible to determine whether the acoustic echo is dominant by using the technique described in Patent Document 1.

  Further, the acoustic echo suppression effect measuring unit 108 compares In (t, f) and LOout (t, f) to determine whether or not the condition shown in the following (Equation 1) is satisfied.

Next, when the acoustic echo suppression effect measuring unit 108 determines that the acoustic echo is dominant in In (t, f) and determines that the condition shown in the above (Equation 1) is satisfied, Considering the instantaneous suppression amount of acoustic echo, P _estm (t, f) is calculated by the _following ( _Equation 2).

In the above (Expression 2), α is a predetermined value, and 0 <α ≦ 1. Here, depending on whether or not the instantaneous suppression amount of acoustic echo is larger than the immediately preceding estimated echo suppression amount P _estm (t−1, f) (frame time one frame before: hereinafter represented by t−1). The value of α may be changed. For example, if (| Lout (t, f) | ² / | In (t, f) | ² ) <P _estm (t−1, f), α is increased (for example, α = 1), and (| Lout ( If t, f) | ² / | In (t, f) | ² ) ≧ P _estm (t−1, f), α is decreased (for example, α = 0.1).

When the acoustic echo suppression effect measuring unit 108 determines that the acoustic echo is not dominant in In (t, f), or determines that the condition shown in (Expression 1) is not satisfied, P _estm (t, f) is calculated by the _following ( _Equation 3) without considering the instantaneous suppression amount of echo.

Then, if the acoustic echo suppression effect measurement unit 108 calculates P _estm (t, f) in the entire frequency band (f = 0 to n−1) for the near-end audio signal at the frame time t, E _linear (t) is calculated by ( _Equation 4). In (Expression 4), n is the number of frequency band divisions in the FFT units 103 and 104.

  Next, details of the output composition unit 109 will be described.

  In the following description, the outputs of the non-linear echo canceller 107 with respect to LOout (t, f) (near-end speech signals after residual echo suppression) are represented as NLOut (t, f), LOut (t, f), and NLOut (t , F), the output of the output synthesis unit 109 is Out (t, f).

  As illustrated in FIG. 1, the output synthesis unit 109 includes an adjustment coefficient calculation unit 1091, output adjustment units 1092 and 1093, and an addition unit 1094.

For each frame, the adjustment coefficient calculation unit 1091 is based on E _linear (t) measured by the acoustic echo suppression effect measurement unit 108 and a preset target value E _target [dB] of the acoustic echo suppression effect. An adjustment coefficient X (t) used for adjusting the sound pressure level of LOOut (t, f) and NLOut (t, f) is calculated.

Here, the target value E _target of the acoustic echo suppression effect may be a fixed value, but since there is no guarantee that the connection destination (the partner terminal) is constant, when the connection destination is changed or in use Even so, it is desirable that the target value E _target of the acoustic echo suppression effect can be changed and set to be variable. For example, the adjustment range of the acoustic echo suppression effect can be set numerically, and the set value is set as the target value E _target of the acoustic echo suppression effect, or the volume adjustment used on the monitor screen of the personal computer is visible. The adjustment range of the acoustic echo suppression effect may be selected and the adjustment range of the acoustic echo suppression effect may be selected, and the selected adjustment content may be digitized to obtain the target value E _target of the acoustic echo suppression effect.

The adjustment coefficient X (t) is calculated by the following (Equation 5) when E _linear (t) is less than a predetermined first threshold value E _min [dB].

When E _linear (t) is equal to or greater than E _min and equal to or less than a predetermined second threshold E _max [dB], X (t) is calculated by the following (Equation 6).

When E _linear (t) is larger than the second threshold E _max, X (t) is calculated by the following (Equation 7).

In the above (Equation 5) to (Equation 7), β is a predetermined value, and 0 <β ≦ 1. Further, E _target <E _min <E _max ≦ 0.

  The output adjustment unit 1092 adjusts the sound pressure level of LOout (t, f) in each frequency band based on X (t) calculated by the adjustment coefficient calculation unit 1091 for each frame. Specifically, the sound pressure level of LOout (t, f) is adjusted by multiplying LOut (t, f) by X (t) / (X (t) +1).

  The output adjustment unit 1093 adjusts the sound pressure level of NLOut (t, f) in each frequency band based on X (t) calculated by the adjustment coefficient calculation unit 1091 for each frame. Specifically, the sound pressure level of NLOut (t, f) is adjusted by multiplying NLOut (t, f) by 1 / (X (t) +1).

  The adder 1094 adds the output of the output adjuster 1092 and the output of the output adjuster 1093 for each frequency band of each frame to calculate Out (t, f). Therefore, Out (t, f) is expressed by the following (Equation 8).

Next, the relationship between E _linear (t) and X (t) will be described.

FIG. 2 is a diagram illustrating an example of a graph representing the relationship between Y (t) and E _linear (t) when Y (t) = 20 log (X (t)). Here, E _target = −45 [dB] and β = 1.

As shown in the drawing, Y (t) increases as the value of E _linear (t) decreases, that is, as the acoustic echo suppression effect of the linear echo canceller 106 increases. As a result, the ratio of LOOut (t, f) to Out (t, f) increases, and the audio distortion of Out (t, f) can be suppressed. Note that the value of Y (t) when E _linear (t) <E _min is fixed to the value of Y (t) when E _linear (t) = E _min . This is because, when X (t) is calculated by the above (Equation 6), if E _linear (t) decreases to a certain value, Y (t) becomes extremely large and many errors are included. Because it becomes. Therefore, as shown in the above (Equation 5), if this certain value is E _min and E _linear (t) <E _min , the value of X (t) is set to E _linear (t) = E ( _min ) is fixed to the value of X (t).

Further, as illustrated, Y (t) decreases as the value of E _linear (t) increases, that is, as the acoustic echo suppression effect of the linear echo canceller unit 106 decreases. As a result, the ratio of NLOut (t, f) in Out (t, f) increases, and the acoustic echo included in Out (t, f) can be suppressed. The value of Y (t) when E _max <E _linear (t) is fixed to the value of Y (t) when E _linear (t) = E _max . This is because, when X (t) is calculated by the above (Equation 6), if E _linear (t) increases to a certain value, Y (t) becomes too small and many errors are included. Because it becomes. Therefore, as shown in the above (Equation 7), when this certain value is E _max and E _max <E _linear (t), the value of X (t) is set to E _linear (t) = The value of X (t) at E _max is fixed.

  The embodiment of the present invention has been described above.

In the present embodiment, the output synthesizer 109 performs the acoustic echo of the near-end speech signal according to the acoustic echo suppression effect E _linear (t) of the linear echo canceller unit 106 measured by the acoustic echo suppression effect measurement unit 108. The output LOout (t, f) of the linear echo canceller unit 106 to be suppressed and the NLOout ( The sound pressure level of t, f) is adjusted, and the two are combined to generate the final output Out (t, f).

Therefore, according to the present embodiment, as shown in FIG. 2, when the acoustic echo suppression effect E _linear (t) of the linear echo canceller unit 106 is large, the linear echo canceller occupying the final output Out (t, f). By increasing the ratio of the output LOOut (t, f) of the unit 106 and reducing the influence of the output NLOOut (t, f) of the nonlinear echo canceller unit 107, the audio distortion of the final output Out (t, f) is reduced. Can be suppressed. When the acoustic echo suppression effect E _linear (t) of the linear echo canceller unit 106 is small, the ratio of the output LOOut (t, f) of the linear echo canceler unit 106 to the final output Out (t, f) is reduced. Thus, by increasing the influence of the output NLOut (t, f) of the nonlinear echo canceller unit 107, the acoustic echo included in the final output Out (t, f) can be suppressed.

  Thus, according to the present embodiment, it is possible to provide an acoustic echo cancellation technique that can suppress voice distortion and enhance the acoustic echo suppression effect.

Further, in the present embodiment, the adjustment coefficient calculation unit 1091 is predetermined within a range where the acoustic echo suppression effect E _linear (t) of the linear echo canceller unit 106 measured by the acoustic echo suppression effect measurement unit 108 is E _min or more. The adjustment coefficient X (t) is calculated so that the ratio of the output LOOut (t, f) of the linear echo canceller 106 to the final output Out (t, f) increases as the target value E _target becomes closer. Therefore, according to the present embodiment, the acoustic echo suppression effect E _linear (t) for the final output Out (t, f) can be brought close to the desired target value E _target while suppressing audio distortion.

Further, in the present embodiment, the acoustic echo suppression effect measuring unit 108 is based on the ratio of the output LOut (t, f) of the linear echo canceller 106 to the input In (t, f) of the linear echo canceller 106. The acoustic echo suppression effect E _linear (t) by the linear echo canceller unit 106 is calculated. Thereby, the acoustic echo suppression effect E _linear (t) by the linear echo canceller 106 can be sequentially measured for each frame.

  In addition, this invention is not limited to the said embodiment, A various deformation | transformation is possible within the range of the summary.

  For example, in the above-described embodiment, the configuration of the echo canceller 1 may be realized in hardware by an integrated logic IC such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA). Alternatively, it may be realized in software by a DSP (Digital Signal Processor). Alternatively, in a general-purpose computer such as an auxiliary storage device such as a CPU, memory, HDD, or DVD-ROM, a modem, and a PC (Personal Computer) equipped with a NIC (Network Interface Card), the CPU stores a predetermined program as an auxiliary storage device. It may be realized by loading the program from the memory into the memory and executing it.

  DESCRIPTION OF SYMBOLS 1: Echo canceller, 101: Far end audio | voice signal input part, 102: Near end audio | voice signal input part, 103: FFT part, 104: FFT part, 105: Pseudo echo generation part, 106: Linear echo canceller part, 107: Nonlinear Echo canceller unit, 108: acoustic echo suppression effect measurement unit, 109: output synthesis unit, 110: IFFT unit, 111: near-end audio signal output unit, 1091: adjustment coefficient calculation unit, 1092: output adjustment unit, 1093: output adjustment Part, 1094: addition part

Claims (5)

An echo canceler that eliminates an acoustic echo of an output audio signal that is an audio signal output from a speaker from an input audio signal that is an audio signal input to a microphone,
Pseudo echo generation means for estimating the acoustic echo of the output audio signal and generating a pseudo echo signal;
Linear echo cancellation means for subtracting the pseudo echo signal from the input audio signal to suppress acoustic echo of the output audio signal contained in the input audio signal;
A residual echo that is a residual component of the acoustic echo included in the input speech signal by changing a sound pressure level for each frequency band for the input speech signal in which the acoustic echo is suppressed by the linear echo canceling unit. A nonlinear echo canceling means for suppressing
Suppression effect measuring means for measuring the suppression effect of the acoustic echo by the linear echo cancellation means;
Based on the measurement result of the suppression effect measuring unit, the input audio signal in which the acoustic echo is suppressed by the linear echo canceling unit and the input audio signal in which the residual echo is suppressed by the nonlinear echo canceling unit An echo canceller comprising: an output synthesizing unit that changes a sound pressure level and synthesizes both. The echo canceller according to claim 1,
The output synthesis means includes
As the suppression effect of the acoustic echo by the linear echo cancellation unit approaches a predetermined target value, the ratio of the input speech signal in which the acoustic echo is suppressed by the linear echo cancellation unit to the output increases. And changing the sound pressure level of each of the input voice signal in which the acoustic echo is suppressed by the linear echo cancellation means and the input voice signal in which the residual echo is suppressed by the nonlinear echo cancellation means. Echo canceller. The echo canceller according to claim 2,
An echo canceller, further comprising target value setting means capable of variably setting the predetermined target value. The echo canceller according to any one of claims 1 to 3,
The suppression effect measuring means includes
An echo canceller that measures the acoustic echo suppression effect of the linear echo cancellation means based on a ratio of the output of the linear echo cancellation to the input of the linear echo cancellation means. An acoustic echo cancellation method for canceling an acoustic echo of an output audio signal, which is an audio signal output from a speaker, by an echo canceller from an input audio signal, which is an audio signal input to a microphone,
The echo canceller is
Estimating the acoustic echo of the output audio signal for each frequency band to generate a pseudo echo signal,
Subtracting the pseudo echo signal from the input voice signal to suppress the acoustic echo of the output voice signal from the input voice signal, and measuring the suppression effect of the acoustic echo,
For the input voice signal in which the acoustic echo is suppressed, the sound pressure level is changed for each frequency band, and the residual echo that is a residual component of the acoustic echo included in the input voice signal is suppressed,
Based on the suppression effect of the acoustic echo, the sound level of each of the input speech signal in which the acoustic echo is suppressed and the input speech signal in which the residual echo is suppressed are changed, and both are synthesized and output. An acoustic echo canceling method characterized by the above.

Images