WO2023092955A1

WO2023092955A1 - Audio signal processing method and apparatus

Info

Publication number: WO2023092955A1
Application number: PCT/CN2022/091811
Authority: WO
Inventors: 张旭; 郑羲光; 李楠; 韩润强; 张晨
Original assignee: 北京达佳互联信息技术有限公司
Priority date: 2021-11-29
Filing date: 2022-05-09
Publication date: 2023-06-01
Also published as: CN114038476A

Abstract

The present disclosure relates to an audio signal processing method and apparatus, and relates to the technical field of signal processing. The audio signal processing method comprises: acquiring a proximal collected audio signal, a distal reference audio signal, and a first proximal audio signal, which is obtained after linear acoustic echo cancellation is performed on the proximal collected audio signal; respectively performing time-frequency transformation on the proximal collected audio signal, the distal reference audio signal and the first proximal audio signal, so as to obtain a frequency domain proximal collected audio signal, a frequency domain distal reference audio signal and a first frequency domain proximal audio signal; performing deep learning noise reduction on the amplitude of the first frequency domain proximal audio signal, so as to obtain a second frequency domain proximal audio signal; performing nonlinear acoustic echo cancellation on the second frequency domain proximal audio signal on the basis of the frequency domain distal reference audio signal, the frequency domain proximal collected audio signal, the first frequency domain proximal audio signal and the second frequency domain proximal audio signal, so as to obtain a frequency domain proximal audio enhanced signal; and performing time-frequency inverse transformation on the frequency domain proximal audio enhanced signal, so as to obtain a proximal audio enhanced signal.

Description

Audio signal processing method and device

Cross References to Related Applications

This application is based on a Chinese patent application with application number 202111433241.1 and a filing date of November 29, 2021, and claims the priority of this Chinese patent application. The entire content of this Chinese patent application is hereby incorporated by reference into this application.

technical field

The present disclosure relates to the technical field of signal processing, and in particular to an audio signal processing method and device.

Background technique

Acoustic Echo Cancellation (AEC) is one of the important technologies in real-time communication, and it is the key point to ensure audio and video experience. Echo cancellation technology refers to the elimination of the far-end signal in the audio signal collected by the near-end microphone, and only the near-end signal is retained. The audio signal collected by the near-end microphone includes the near-end signal and the far-end signal through the near-end The signal played by the speaker. Echo cancellation techniques generally include linear echo cancellation and nonlinear echo cancellation.

Contents of the invention

The disclosure provides an audio signal processing method and device.

According to the first aspect of an embodiment of the present disclosure, an audio signal processing method is provided, including: acquiring a near-end collected audio signal, a far-end reference audio signal, and performing linear echo cancellation on the near-end collected audio signal. A near-end audio signal; time-frequency conversion is performed on the near-end collected audio signal, the far-end reference audio signal and the first near-end audio signal respectively, to obtain the frequency-domain near-end collected audio signal, the frequency-domain far-end A reference audio signal and a first frequency domain near-end audio signal; performing deep learning noise reduction on the magnitude of the first frequency domain near-end audio signal to obtain a second frequency domain near-end audio signal; based on the frequency domain far-end reference audio signal, the near-end audio signal in the frequency domain, the near-end audio signal in the first frequency domain, and the near-end audio signal in the second frequency domain, performing nonlinear echo on the near-end audio signal in the second frequency domain elimination to obtain a frequency-domain near-end audio enhancement signal; performing time-frequency inverse transform on the frequency-domain near-end audio enhancement signal to obtain a near-end audio enhancement signal.

In some embodiments, performing deep learning noise reduction on the amplitude of the first frequency-domain near-end audio signal to obtain the second frequency-domain near-end audio signal includes: using a trained noise reduction neural network model to The magnitude of the near-end audio signal in the first frequency domain is subjected to deep learning noise reduction to obtain the magnitude of the near-end audio signal in the second frequency domain; according to the magnitude of the near-end audio signal in the second frequency domain and the first Phase of the near-end audio signal in the frequency domain to obtain the second near-end audio signal in the frequency domain.

In some embodiments, the trained noise reduction neural network model performs deep learning and noise reduction on the amplitude of the first frequency-domain near-end audio signal to obtain the amplitude of the second frequency-domain near-end audio signal , comprising: inputting the magnitude of the first frequency-domain near-end audio signal into the trained noise reduction neural network model to obtain a first signal magnitude ratio, wherein the first signal magnitude ratio is the second The predicted value of the ratio of the amplitude of the near-end audio signal in the frequency domain to the magnitude of the first frequency-domain near-end audio signal; according to the first signal amplitude ratio and the magnitude of the first frequency-domain near-end audio signal, obtain The magnitude of the second frequency-domain near-end audio signal, wherein the magnitude of the second frequency-domain near-end audio signal is the product of the first signal magnitude ratio and the magnitude of the first frequency-domain near-end audio signal .

In some embodiments, the frequency-domain far-end reference audio signal, the frequency-domain near-end audio signal, the first frequency-domain near-end audio signal and the second frequency-domain near-end audio signal , performing nonlinear echo cancellation on the second frequency-domain near-end audio signal to obtain a frequency-domain near-end audio enhancement signal, comprising: respectively combining the frequency-domain far-end reference audio signal with the frequency-domain near-end acquisition audio signal Perform correlation with the second frequency near-end audio signal on each frequency band to obtain a second signal amplitude ratio of each frequency band; according to the second signal amplitude ratio, the first frequency domain near-end audio signal and the second For the near-end audio signal in the frequency domain, nonlinear echo cancellation is performed on the second near-end audio signal in the frequency domain to obtain an enhanced near-end audio signal in the frequency domain.

In some embodiments, according to the second signal amplitude ratio, the first frequency-domain near-end audio signal and the second frequency-domain near-end audio signal, the second frequency-domain near-end audio signal Perform nonlinear echo cancellation to obtain a frequency-domain near-end audio enhancement signal, including: obtaining the product of the second signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal as a reference amplitude; obtaining the reference amplitude and The minimum value of the amplitude of the second frequency-domain near-end audio signal is used as the amplitude of the frequency-domain near-end audio enhancement signal; according to the amplitude of the frequency-domain near-end audio enhancement signal and the first frequency-domain near-end The phase of the end audio signal is obtained to obtain the near-end audio enhancement signal in the frequency domain.

In some embodiments, the second signal amplitude ratio is obtained by the following formula:

Mask(n,k)=min{1-RCr(n,k),1-RY _p r(n,k)};

Wherein, Mask(n,k) is the second signal amplitude ratio, RCr(n,k) is the cross-correlation coefficient between the frequency-domain far-end reference audio signal and the frequency-domain near-end acquisition audio signal, _RYp r(n,k) is the cross-correlation coefficient between the frequency domain far-end reference audio signal and the second frequency domain near-end audio signal, n is the number of frame sequences, k is the number of center frequency sequences, 0<n≤N , 0<k≤K, N is the total number of frames, K is the total number of frequency bands.

According to a second aspect of an embodiment of the present disclosure, there is provided an audio signal processing device, including: a signal acquisition unit configured to: acquire a near-end collected audio signal, a far-end reference audio signal, and perform an operation on the near-end collected audio signal The first near-end audio signal obtained after performing linear echo cancellation; the frequency domain transformation unit is configured to: respectively perform Time-frequency transformation to obtain the frequency-domain near-end audio signal, the frequency-domain far-end reference audio signal and the first frequency-domain near-end audio signal; the deep noise reduction unit is configured to: the first frequency-domain near-end audio signal The magnitude of the deep learning noise reduction is carried out to obtain the second frequency domain near-end audio signal; the non-linear elimination unit is configured to: based on the frequency domain far-end reference audio signal, the frequency domain near-end acquisition audio signal, the The first frequency-domain near-end audio signal and the second frequency-domain near-end audio signal, performing nonlinear echo cancellation on the second frequency-domain near-end audio signal to obtain a frequency-domain near-end audio enhancement signal; a time-domain transformation unit , is configured to: perform a time-frequency inverse transform on the frequency-domain near-end audio enhancement signal to obtain a near-end audio enhancement signal.

In some embodiments, the deep noise reduction unit is configured to: use a trained noise reduction neural network model to perform deep learning and noise reduction on the amplitude of the near-end audio signal in the first frequency domain to obtain the second frequency domain Amplitude of the near-end audio signal; the second frequency-domain near-end audio signal is obtained according to the magnitude of the second frequency-domain near-end audio signal and the phase of the first frequency-domain near-end audio signal.

In some embodiments, the deep noise reduction unit is configured to: input the magnitude of the first frequency-domain near-end audio signal into the trained noise reduction neural network model to obtain a first signal magnitude ratio, wherein the The first signal amplitude ratio is a predicted value of the ratio of the amplitude of the second frequency-domain near-end audio signal to the amplitude of the first frequency-domain near-end audio signal; according to the first signal amplitude ratio and the first The magnitude of a near-end audio signal in the frequency domain is obtained by obtaining the magnitude of the second frequency-domain near-end audio signal, wherein the magnitude of the second frequency-domain near-end audio signal is the first signal magnitude ratio and the first A product of the amplitudes of the near-end audio signal in the frequency domain.

In some embodiments, the non-linear elimination unit is configured to: correlate the frequency-domain far-end reference audio signal with the frequency-domain near-end acquisition audio signal and the second-frequency near-end audio signal in respective frequency bands , to obtain the second signal amplitude ratio of each frequency band; according to the second signal amplitude ratio, the first frequency domain near-end audio signal and the second frequency domain near-end audio signal, the second frequency domain near-end audio signal The non-linear echo cancellation is performed on the end audio signal to obtain the near-end audio enhancement signal in the frequency domain.

In some embodiments, the non-linear elimination unit is configured to: acquire the product of the second signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal as a reference amplitude; acquire the reference amplitude and the first The minimum value of the amplitudes of the two frequency-domain near-end audio signals is used as the amplitude of the frequency-domain near-end audio enhancement signal; according to the amplitude of the frequency-domain near-end audio enhancement signal and the first frequency-domain near-end audio signal The phase of the near-end audio enhancement signal in the frequency domain is obtained.

Mask(n,k)=min{1-RCr(n,k),1-RY _p r(n,k)};

According to a third aspect of the embodiments of the present disclosure, there is provided an electronic device, including: at least one processor; at least one memory storing computer-executable instructions, wherein the computer-executable instructions are executed by the at least one processor , prompting the at least one processor to execute the audio signal processing method according to the present disclosure.

According to a fourth aspect of the embodiments of the present disclosure, there is provided a non-volatile computer-readable storage medium, when instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is prompted to execute An audio signal processing method according to the present disclosure.

According to a fifth aspect of the embodiments of the present disclosure, there is provided a computer program product including computer instructions, and when the computer instructions are executed by at least one processor, the audio signal processing method according to the present disclosure is implemented.

According to the audio signal processing method and device of the present disclosure, linear echo cancellation is first performed on the near-end collected audio signal, and then deep learning noise reduction is performed on it, and then nonlinear echo cancellation is performed. The echo signal in the end reference audio signal is eliminated to obtain the final near-end audio enhancement signal, and the combination of deep learning noise reduction and echo cancellation makes full use of the good performance of deep learning noise reduction. The combination of traditional noise reduction technology and echo cancellation can get better echo cancellation effect and noise reduction effect, and bring about the improvement of sound quality.

In addition, according to the audio signal processing method and device of the present disclosure, deep learning can be used for noise reduction to deal with situations that are difficult to solve in traditional noise reduction, such as the elimination of quasi-stationary or non-stationary noise and the traditional noise reduction that is not effective in eliminating noise. Condition.

In addition, according to the audio signal processing method and device of the present disclosure, in the process of nonlinear echo cancellation, by obtaining the signal amplitude ratio based on the cross-correlation coefficient, and finally obtaining the near-end audio enhancement signal in the frequency domain, the echo and noise can be eliminated at the same time.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.

Description of drawings

The accompanying drawings here are incorporated into the specification and constitute a part of the specification, show embodiments consistent with the disclosure, and are used together with the description to explain the principle of the disclosure, and do not constitute an improper limitation of the disclosure.

FIG. 1 is an overall block diagram illustrating an audio signal processing method according to an exemplary embodiment of the present disclosure.

FIG. 2 is a flowchart illustrating an audio signal processing method according to an exemplary embodiment of the present disclosure.

FIG. 3 is a block diagram illustrating an audio signal processing device according to an exemplary embodiment of the present disclosure.

FIG. 4 is a block diagram illustrating an electronic device 400 according to an exemplary embodiment of the present disclosure.

Detailed ways

In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

It should be noted that the terms "first" and "second" in the specification and claims of the present disclosure and the above drawings are used to distinguish similar objects, but not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein can be practiced in sequences other than those illustrated or described herein. The implementations described in the following examples do not represent all implementations consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with aspects of the present disclosure as recited in the appended claims.

What needs to be explained here is that "at least one of several items" appearing in this disclosure all means to include "any one of the several items", "a combination of any of the several items", The three categories of "the whole of the several items" are juxtaposed. For example, "including at least one of A and B" includes the following three parallel situations: (1) including A; (2) including B; (3) including A and B. Another example is "execute at least one of step 1 and step 2", which means the following three parallel situations: (1) execute step 1; (2) execute step 2; (3) execute step 1 and step 2.

In real-time communication, audio signals contain noise as well as echo. Based on this, Acoustic Echo Cancellation (AEC) has become one of the important technologies in real-time communication, and it is the key point to ensure audio and video experience. Echo cancellation technology refers to the elimination of the far-end signal in the audio signal collected by the near-end microphone, and only the near-end signal is retained. The audio signal collected by the near-end microphone includes the near-end signal and the far-end signal through the near-end The audio signal played by the speaker. It should be noted that the near end may be referred to as the local end, and the far end may be referred to as the other end. Echo cancellation techniques generally include linear echo cancellation and nonlinear echo cancellation.

Linear echo cancellation can eliminate far-end signals through adaptive filtering, but often there will be residual far-end signals. The combination of linear echo cancellation and nonlinear echo cancellation will improve the cancellation effect of the far-end signal. But usually, the signal-to-noise ratio of the signal collected at the near end will affect the effect of nonlinear echo cancellation. In related technologies, the audio signal that has undergone linear echo cancellation is noise-reduced using a traditional method such as Wiener filtering to obtain a relatively clean audio signal, and then non-linear echo cancellation is performed for correlation processing. However, the audio signal processed by this method is limited by the effect of traditional noise reduction technology, and the noise reduction effect is not good, especially in the case of non-stationary noise, which affects the echo cancellation effect and causes poor sound quality.

Combine echo cancellation with traditional noise reduction such as filtering to achieve the effects of noise reduction and echo cancellation, but the audio signal processed by this method has poor echo cancellation and noise reduction effects and poor sound quality.

The present disclosure proposes an audio signal processing method and device, which first performs linear echo cancellation on the near-end audio signal, then performs deep learning noise reduction on it, and then performs nonlinear echo cancellation to obtain the final near-end audio enhancement signal. The combination of deep learning noise reduction and echo cancellation, compared with the combination of traditional noise reduction technology and echo cancellation used in related technologies, can obtain better echo cancellation and noise reduction effects, resulting in improved sound quality.

Hereinafter, the audio signal processing method and device according to the present disclosure will be described in detail with reference to FIGS. 1 to 4 .

FIG. 1 is an overall block diagram illustrating an audio signal processing method according to an exemplary embodiment of the present disclosure. Referring to Figure 1, the near-end collected audio signal can be subjected to linear echo cancellation to obtain the first near-end audio signal, and the first frequency-domain near-end audio signal can be subjected to deep learning noise reduction to obtain the second frequency-domain near-end audio signal, Non-linear echo cancellation processing can be performed on the second frequency-domain near-end audio signal to obtain a frequency-domain near-end audio enhancement signal. It should be noted that, in this process, the far-end reference The echo signal in the audio signal is eliminated.

After knowing the overall framework of the exemplary embodiment of the present disclosure, the audio signal processing method of the exemplary embodiment of the present disclosure will be described below through specific method steps.

FIG. 2 is a flowchart illustrating an audio signal processing method according to an exemplary embodiment of the present disclosure. The audio signal processing method provided in the embodiment of the present disclosure includes: Step 201, acquiring a near-end audio signal, a far-end reference audio signal, and a first near-end audio signal obtained by performing linear echo cancellation on the near-end audio signal ; Step 202, time-frequency conversion is performed on the near-end audio signal, the far-end reference audio signal and the first near-end audio signal respectively, to obtain the near-end audio signal in the frequency domain, the far-end reference audio signal in the frequency domain and the first frequency domain Near-end audio signal; Step 203, carry out deep learning noise reduction to the amplitude of the first frequency domain near-end audio signal, obtain the second frequency domain near-end audio signal; Step 204, based on frequency domain far-end reference audio signal, frequency domain near-end The terminal collects the audio signal, the near-end audio signal in the first frequency domain and the near-end audio signal in the second frequency domain, performs nonlinear echo cancellation on the near-end audio signal in the second frequency domain, and obtains the near-end audio enhancement signal in the frequency domain; step 205, Time-frequency inverse transform is performed on the near-end audio enhancement signal in the frequency domain to obtain the near-end audio enhancement signal. Referring to FIG. 2 , in step 201 , a near-end collected audio signal, a far-end reference audio signal, and a first near-end audio signal obtained after performing linear echo cancellation on the near-end collected audio signal can be obtained.

According to an exemplary embodiment of the present disclosure, the near-end collected audio signal is an audio signal collected by a near-end microphone, and the far-end reference audio signal is an audio signal transmitted from a far end that is not played by a near-end speaker.

The audio signal collected by the near-end microphone may include, but not limited to, the near-end user voice signal collected by the near-end microphone and the audio collected by the near-end microphone and transmitted from the far end through the network link and played through the near-end speaker Signal.

In step 202, time-frequency transformation may be performed on the near-end collected audio signal, the far-end reference audio signal and the first near-end audio signal respectively, to obtain the frequency-domain near-end collected audio signal, the frequency-domain far-end reference audio signal and the first frequency domain domain near-end audio signal.

According to an exemplary embodiment of the present disclosure, the time-frequency transform may be, but not limited to, Short-Time Fourier Transform (Short-Time Fourier Transform, STFT). The STFT is taken as an example for description below.

According to an exemplary embodiment of the present disclosure, the near-end acquisition audio signal, the far-end reference audio signal and the first near-end audio signal with a time length of T may be expressed as c(t) and r(t) respectively in the time domain and y(t), where t is time, 0<t≤T. After short-time Fourier transform, c(t), r(t) and y(t) can be expressed in the frequency domain by the following formulas (1)-(3):

C(n,k)=STFT(c(t));(1)

R(n,k)=STFT(r(t));(2)

Y(n,k)=STFT(y(t));(3)

Among them, c(t) is the near-end acquisition audio signal, r(t) is the far-end reference audio signal, y(t) is the first near-end audio signal, and STFT(x) represents the short-time Fourier transform of x , C(n,k) is the frequency-domain near-end audio signal, R(n,k) is the frequency-domain far-end reference audio signal, Y(n,k) is the first frequency-domain near-end audio signal, and n is the frame Number of sequences, k is the number of center frequency sequences, 0<n≤N, 0<k≤K, N is the total number of frames, K is the total number of frequency bands.

In step 203, the amplitude of the near-end audio signal in the first frequency domain can be denoised by deep learning to obtain the near-end audio signal in the second frequency domain.

According to an exemplary embodiment of the present disclosure, deep learning noise reduction can be implemented by using a neural network model to perform noise reduction. In this case: firstly, the first frequency domain near-end audio signal can be processed through a trained noise reduction neural network model The amplitude of the deep learning noise reduction is performed to obtain the amplitude of the near-end audio signal in the second frequency domain. Then, the second frequency-domain near-end audio signal can be obtained according to the amplitude of the second frequency-domain near-end audio signal and the phase of the first frequency-domain near-end audio signal.

According to an exemplary embodiment of the present disclosure, first, the magnitude of the first frequency-domain near-end audio signal can be input into the trained noise reduction neural network model to obtain the first signal magnitude ratio, wherein the first signal magnitude ratio is the second A predicted value of a ratio of the amplitude of the frequency-domain near-end audio signal to the amplitude of the first frequency-domain near-end audio signal. Then the amplitude of the near-end audio signal in the second frequency domain can be obtained according to the amplitude ratio of the first signal amplitude and the amplitude of the near-end audio signal in the first frequency domain, wherein the amplitude of the near-end audio signal in the second frequency domain is the first signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal. For example, the trained noise reduction neural network model may be a neural network model based on deep learning (Deep Learning), including, but not limited to, a convolutional neural network model.

According to an exemplary embodiment of the present disclosure, the training of the noise reduction neural network model to be trained is completed by training the training data set, wherein the training data set may include an amplitude data set with a noisy frequency signal.

According to an exemplary embodiment of the present disclosure, the second frequency-domain near-end audio signal can be obtained by the following equation (4):

Y(n,k) _p =PrediCt(MagY(n,k))*Phase(Y(n,k));(4)

Wherein, Y(n,k) _p is the near-end audio signal in the second frequency domain, n is the number of frame sequences, k is the number of center frequency sequences, 0<n≤N, 0<k≤K, N is the total number of frames, K is the total number of frequency bands, Y(n,k) is the near-end audio signal in the first frequency domain, MagY(n,k) is the amplitude of the near-end audio signal in the first frequency domain, Predict(MagY(n,k)) represents Perform deep learning noise reduction on the amplitude of the near-end audio signal in the first frequency domain, and Phase(Y(n,k)) is the phase of the near-end audio signal in the first frequency domain.

Returning to Fig. 2, in step 204, based on the far-end reference audio signal in the frequency domain, the near-end audio signal in the frequency domain, the near-end audio signal in the first frequency domain and the near-end audio signal in the second frequency domain, the second frequency domain The near-end audio signal is subjected to nonlinear echo cancellation to obtain a near-end audio enhancement signal in the frequency domain.

According to an exemplary embodiment of the present disclosure, firstly, the frequency-domain far-end reference audio signal can be correlated with the frequency-domain near-end collected audio signal and the second-frequency near-end audio signal in each frequency band to obtain the second signal amplitude ratio. Then, according to the second signal amplitude ratio, the first frequency-domain near-end audio signal and the second frequency-domain near-end audio signal, nonlinear echo cancellation is performed on the second frequency-domain near-end audio signal to obtain a frequency-domain near-end audio enhancement signal .

According to an exemplary embodiment of the present disclosure, first, the cross-correlation coefficient of the frequency-domain far-end reference audio signal and the frequency-domain near-end acquisition audio signal can be obtained, and the frequency-domain far-end reference audio signal and the second frequency-domain near-end audio signal The correlation coefficient of . The second signal amplitude ratios for the respective frequency bands can then be obtained based on these two cross-correlation coefficients.

For example, the cross-correlation coefficient of the frequency-domain far-end reference audio signal and the frequency-domain near-end acquisition audio signal can be obtained by the following formula (5):

RCr(n,k)=Xcorr(R(n,k),C(n,k));(5)

Among them, RCr(n,k) is the cross-correlation coefficient of the far-end reference audio signal in the frequency domain and the near-end acquisition audio signal in the frequency domain, and Xcorr(a,b) represents the correlation between a and b in each frame and each frequency band through the cross-correlation function Find the cross-correlation to obtain the cross-correlation coefficients of a and b in each frequency band, R(n,k) is the far-end reference audio signal in the frequency domain, C(n,k) is the near-end audio signal in the frequency domain, and n is the frame sequence number, k is the number of center frequency sequences, 0<n≤N, 0<k≤K, N is the total number of frames, and K is the total number of frequency bands.

For example, the cross-correlation coefficient of the far-end reference audio signal in the frequency domain and the near-end audio signal in the second frequency domain can be obtained by the following formula (6):

RY _p r(n,k)=Xcorr(R(n,k),Y(n,k) _p );(6)

Among them, RY _p r(n,k) is the cross-correlation coefficient of the far-end reference audio signal in the frequency domain and the near-end audio signal in the second frequency domain, and Xcorr(a,b) represents the correlation between a and b in each frame through the cross-correlation function Calculate the cross-correlation with each frequency band to obtain the cross-correlation coefficient of a and b in each frequency band, R(n,k) is the far-end reference audio signal in the frequency domain, and Y(n,k) _p is the near-end audio signal in the second frequency domain , n is the number of frame sequences, k is the number of center frequency sequences, 0<n≤N, 0<k≤K, N is the total number of frames, and K is the total number of frequency bands.

For example, the second signal amplitude ratio is obtained by the following formula (7):

Mask(n,k)=min{1-RCr(n,k),1-RY _p r(n,k)}; (7)

Among them, Mask(n,k) is the second signal amplitude ratio, RCr(n,k) is the cross-correlation coefficient between the frequency domain far-end reference audio signal and the frequency domain near-end acquisition audio signal, RY _p r(n,k) is the cross-correlation coefficient between the far-end reference audio signal in the frequency domain and the near-end audio signal in the second frequency domain, n is the number of frame sequences, k is the number of center frequency sequences, 0<n≤N, 0<k≤K, N is the total The number of frames, K is the total number of frequency bands.

According to an exemplary embodiment of the present disclosure, first, the product of the second signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal may be acquired as the reference amplitude. Then the minimum value of the reference amplitude and the amplitude of the second frequency-domain near-end audio signal may be obtained as the amplitude of the frequency-domain near-end audio enhancement signal. Finally, the frequency-domain near-end audio enhancement signal can be obtained according to the amplitude of the frequency-domain near-end audio enhancement signal and the phase of the first frequency-domain near-end audio signal.

For example, the near-end audio enhancement signal in the frequency domain can be obtained by the following formula (8):

Y(n,k) _out =min{MagY(n,k)*Mask(n,k),MagY(n,k) _p }*Phase(Y(n,k));(8)

Among them, Y(n,k) _out is the near-end audio enhancement signal in the frequency domain, n is the number of frame sequences, k is the number of center frequency sequences, 0<n≤N, 0<k≤K, N is the total number of frames, K is the total number of frequency bands, Y(n,k) is the near-end audio signal in the first frequency domain, MagY(n,k) is the amplitude of the near-end audio signal in the first frequency domain, and Mask(n,k) is the second signal amplitude ratio, MagY(n,k) _p is the amplitude of the near-end audio signal in the second frequency domain, Y(n,k) _p is the near-end audio signal in the second frequency domain, and Phase(Y(n,k)) is the first The phase of the near-end audio signal in the frequency domain.

Returning to FIG. 2 , in step 205 , time-frequency inverse transform may be performed on the near-end audio enhancement signal in the frequency domain to obtain the near-end audio enhancement signal.

According to an exemplary embodiment of the present disclosure, the inverse time-frequency transform may be, but not limited to, Inverse Short-Time Fourier Transform (ISTFT). In this case, the near-end audio enhancement signal can be obtained by the following formula (9):

y(t) _out = ISTFT(Y(n,k) _out ); (9)

Among them, y(t) _out is the near-end audio enhancement signal, ISTFT(x) represents the short-time inverse Fourier transform of x, Y(n,k) _out is the frequency domain near-end audio enhancement signal, t is time, 0<t≤T, T is the time length, n is the number of frame sequences, k is the number of center frequency sequences, 0<n≤N, 0<k≤K, N is the total number of frames, K is the total number of frequency bands.

FIG. 3 is a block diagram illustrating an audio signal processing device according to an exemplary embodiment of the present disclosure. Referring to FIG. 3 , an audio signal processing device 300 according to an exemplary embodiment of the present disclosure may include a signal acquisition unit 301 , a frequency domain transformation unit 302 , a depth noise reduction unit 303 , a non-linear elimination unit 304 and a time domain transformation unit 305 .

The signal obtaining unit 301 may obtain a near-end collected audio signal, a far-end reference audio signal, and a first near-end audio signal obtained by performing linear echo cancellation on the near-end collected audio signal.

The frequency domain conversion unit 302 can perform time-frequency conversion on the near-end collected audio signal, the far-end reference audio signal and the first near-end audio signal, respectively, to obtain the frequency-domain near-end collected audio signal, the frequency-domain far-end reference audio signal and the first Frequency domain near-end audio signal.

According to an exemplary embodiment of the present disclosure, the near-end acquisition audio signal, the far-end reference audio signal and the first near-end audio signal with a time length of T may be expressed as c(t) and r(t) respectively in the time domain and y(t), where t is time, 0<t≤T. After short-time Fourier transform, c(t), r(t) and y(t) can be expressed by the above formulas (1)-(3) in the frequency domain.

The deep noise reduction unit 303 may perform deep learning and noise reduction on the magnitude of the first frequency-domain near-end audio signal to obtain a second frequency-domain near-end audio signal.

According to an exemplary embodiment of the present disclosure, deep learning denoising can be implemented by using a neural network model to perform denoising. The amplitude of the near-end audio signal in the frequency domain is used for deep learning noise reduction, and the amplitude of the near-end audio signal in the second frequency domain is obtained. The deep noise reduction unit 303 can then obtain the second frequency-domain near-end audio signal according to the amplitude of the second frequency-domain near-end audio signal and the phase of the first frequency-domain near-end audio signal.

According to an exemplary embodiment of the present disclosure, the deep noise reduction unit 303 may first input the amplitude of the first frequency domain near-end audio signal into the trained noise reduction neural network model to obtain the first signal amplitude ratio, wherein the first signal The amplitude ratio is a predicted value of a ratio of the amplitude of the near-end audio signal in the second frequency domain to the amplitude of the near-end audio signal in the first frequency domain. The depth noise reduction unit 303 can then obtain the magnitude of the second frequency domain near-end audio signal according to the first signal magnitude ratio and the magnitude of the first frequency domain near-end audio signal, wherein the magnitude of the second frequency domain near-end audio signal is The product of the first signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal. For example, the trained noise reduction neural network model may be a neural network model based on deep learning (Deep Learning), including, but not limited to, a convolutional neural network model.

According to an exemplary embodiment of the present disclosure, the second frequency-domain near-end audio signal can be obtained by the above formula (4).

Returning to FIG. 3 , the non-linear elimination unit 304 may perform a second-frequency The non-linear echo cancellation is performed on the near-end audio signal in the frequency domain to obtain the near-end audio enhancement signal in the frequency domain.

According to an exemplary embodiment of the present disclosure, the non-linear elimination unit 304 may first correlate the frequency-domain far-end reference audio signal with the frequency-domain near-end collected audio signal and the second-frequency near-end audio signal in each frequency band to obtain Second signal amplitude ratios for respective frequency bands. The nonlinear elimination unit 304 can then perform nonlinear echo cancellation on the second frequency-domain near-end audio signal according to the second signal amplitude ratio, the first frequency-domain near-end audio signal, and the second frequency-domain near-end audio signal to obtain a frequency domain Near-end audio enhancement signal.

According to an exemplary embodiment of the present disclosure, the nonlinear elimination unit 304 may first obtain the cross-correlation coefficient between the frequency-domain far-end reference audio signal and the frequency-domain near-end acquisition audio signal, and the frequency-domain far-end reference audio signal and the second frequency The cross-correlation coefficient of the domain near-end audio signal. The non-linear elimination unit 304 can then obtain the second signal amplitude ratio of each frequency band based on the two correlation coefficients.

For example, the cross-correlation coefficient between the frequency-domain far-end reference audio signal and the frequency-domain near-end collected audio signal can be obtained through the above formula (5).

For example, the cross-correlation coefficient between the far-end reference audio signal in the frequency domain and the near-end audio signal in the second frequency domain may be obtained through the above formula (6).

For example, the second signal amplitude ratio is obtained through the above formula (7). According to an exemplary embodiment of the present disclosure, the nonlinear elimination unit 304 may first obtain a product of the second signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal as the reference amplitude. The non-linear elimination unit 304 may then obtain the minimum value of the reference amplitude and the amplitude of the second frequency-domain near-end audio signal as the amplitude of the frequency-domain near-end audio enhancement signal. Finally, the non-linear elimination unit 304 can obtain the frequency-domain near-end audio enhancement signal according to the amplitude of the frequency-domain near-end audio enhancement signal and the phase of the first frequency-domain near-end audio signal.

For example, the near-end audio enhancement signal in the frequency domain can be obtained through the above formula (8).

Returning to FIG. 3 , the time-domain transformation unit 305 may perform time-frequency inverse transformation on the near-end audio enhancement signal in the frequency domain to obtain the near-end audio enhancement signal.

According to an exemplary embodiment of the present disclosure, the inverse time-frequency transform may be, but not limited to, Inverse Short-Time Fourier Transform (ISTFT). In this case, the near-end audio enhancement signal can be obtained through the above formula (9).

4, the electronic device 400 includes at least one memory 401 and at least one processor 402, the at least one memory 401 stores a set of computer-executable instructions, when the set of computer-executable instructions is executed by the at least one processor 402, the execution An audio signal processing method according to an exemplary embodiment of the present disclosure.

As an example, the electronic device 400 may be a PC computer, a tablet device, a personal digital assistant, a smart phone, or other devices capable of executing the above-mentioned set of instructions. Here, the electronic device 400 is not necessarily a single electronic device, but may also be any assembly of devices or circuits capable of individually or jointly executing the above-mentioned instructions (or instruction sets). The electronic device 400 may also be part of an integrated control system or system manager, or may be configured as a portable electronic device that interfaces locally or remotely (eg, via wireless transmission).

In electronic device 400, processor 402 may include a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a special purpose processor system, a microcontroller, or a microprocessor. By way of example and not limitation, a processor may also include analog processors, digital processors, microprocessors, multi-core processors, processor arrays, network processors, and the like.

The processor 402 can execute instructions or codes stored in the memory 401, wherein the memory 401 can also store data. Instructions and data may also be sent and received over the network via the network interface device, which may employ any known transmission protocol.

The memory 401 can be integrated with the processor 402, for example, RAM or flash memory is arranged in an integrated circuit microprocessor or the like. Additionally, memory 401 may comprise a separate device, such as an external disk drive, storage array, or any other storage device usable by the database system. The memory 401 and the processor 402 may be operatively coupled, or may communicate with each other, such as through an I/O port, network connection, etc., such that the processor 402 can read files stored in the memory.

In addition, the electronic device 400 may further include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, mouse, touch input device, etc.). All components of the electronic device 400 may be connected to each other via a bus and/or a network.

According to an exemplary embodiment of the present disclosure, there may also be provided a non-transitory computer-readable storage medium storing instructions, wherein, when the instructions are executed by at least one processor, at least one processor is caused to execute an example according to the present disclosure. The audio signal processing method of the exemplary embodiment. Examples of computer readable storage media herein include: Read Only Memory (ROM), Random Access Programmable Read Only Memory (PROM), Electrically Erasable Programmable Read Only Memory (EEPROM), Random Access Memory (RAM) , Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Flash Memory, Non-volatile Memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM , DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-R LTH, BD-RE, Blu-ray or Optical Memory, Hard Disk Drive (HDD), Solid State Hard disks (SSD), memory cards (such as MultiMediaCards, Secure Digital (SD) or Extreme Digital (XD) cards), magnetic tapes, floppy disks, magneto-optical data storage devices, optical data storage devices, hard disks, solid-state disks, and any other means configured to store a computer program and any associated data, data files and data structures in a non-transitory manner and to provide said computer program and any associated data, data files and data structures to the processor or the computer to enable the processor or the computer to execute the computer program. The computer program in the above-mentioned computer-readable storage medium can run in an environment deployed in computer equipment such as a client, a host, an agent device, a server, etc. In addition, in one example, the computer program and any associated data and data files and data structures are distributed over network-connected computer systems so that the computer programs and any associated data, data files and data structures are stored, accessed and executed in a distributed fashion by one or more processors or computers.

According to an exemplary embodiment of the present disclosure, there may also be provided a computer program product, instructions in the computer program product may be executed by a processor of a computer device to implement the audio signal processing method according to the exemplary embodiment of the present disclosure.

According to the audio signal processing method and device of the present disclosure, linear echo cancellation is first performed on the near-end collected audio signal, and then deep learning noise reduction is performed on it, and then nonlinear echo cancellation is performed. The echo signal in the end reference audio signal is eliminated to obtain the final near-end audio enhancement signal, and the combination of deep learning noise reduction and echo cancellation makes full use of the good performance of deep learning noise reduction. The combination of traditional noise reduction technology and echo cancellation can get better echo cancellation and noise reduction, and improve sound quality.

All the embodiments of the present disclosure can be implemented independently or in combination with other embodiments, which are all regarded as the scope of protection required by the present disclosure.

Claims

An audio signal processing method, comprising:

Acquiring a near-end audio signal, a far-end reference audio signal, and a first near-end audio signal obtained by performing linear echo cancellation on the near-end audio signal;

performing time-frequency transformation on the near-end collected audio signal, the far-end reference audio signal and the first near-end audio signal respectively, to obtain the frequency-domain near-end collected audio signal, the frequency-domain far-end reference audio signal and the first frequency domain near-end audio signal;

performing deep learning noise reduction on the amplitude of the first frequency-domain near-end audio signal to obtain a second frequency-domain near-end audio signal;

Based on the frequency-domain far-end reference audio signal, the frequency-domain near-end collected audio signal, the first frequency-domain near-end audio signal, and the second frequency-domain near-end audio signal, the second frequency domain Non-linear echo cancellation is performed on the near-end audio signal to obtain a near-end audio enhancement signal in the frequency domain;

Inverse time-frequency transform is performed on the near-end audio enhancement signal in the frequency domain to obtain the near-end audio enhancement signal.
The audio signal processing method according to claim 1, wherein said performing deep learning noise reduction on the magnitude of the first frequency-domain near-end audio signal to obtain a second frequency-domain near-end audio signal, comprising:

Through the trained noise reduction neural network model, the magnitude of the near-end audio signal in the first frequency domain is subjected to deep learning and noise reduction to obtain the magnitude of the near-end audio signal in the second frequency domain;

The second frequency-domain near-end audio signal is obtained according to the amplitude of the second frequency-domain near-end audio signal and the phase of the first frequency-domain near-end audio signal.
The audio signal processing method according to claim 2, wherein, the amplitude of the near-end audio signal in the first frequency domain is subjected to deep learning and noise reduction through the trained noise reduction neural network model to obtain the second The magnitude of the near-end audio signal in the frequency domain, including:

Inputting the magnitude of the near-end audio signal in the first frequency domain into the trained noise reduction neural network model to obtain a first signal magnitude ratio, wherein the first signal magnitude ratio is equal to the second frequency domain near-end audio signal A predicted value of the ratio of the amplitude of the end audio signal to the amplitude of the first frequency domain near end audio signal;

According to the first signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal, the amplitude of the second frequency-domain near-end audio signal is obtained, wherein the amplitude of the second frequency-domain near-end audio signal is the product of the first signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal.
The audio signal processing method according to claim 1, wherein said frequency-domain far-end reference audio signal based on said frequency-domain near-end acquisition audio signal, said first frequency-domain near-end audio signal and said A second frequency-domain near-end audio signal, performing nonlinear echo cancellation on the second frequency-domain near-end audio signal to obtain a frequency-domain near-end audio enhancement signal, including:

Correlating the frequency-domain far-end reference audio signal with the frequency-domain near-end acquisition audio signal and the second-frequency near-end audio signal in each frequency band to obtain a second signal amplitude ratio of each frequency band;

According to the second signal amplitude ratio, the first frequency-domain near-end audio signal, and the second frequency-domain near-end audio signal, nonlinear echo cancellation is performed on the second frequency-domain near-end audio signal to obtain a frequency Domain near-end audio enhancement signal.
The audio signal processing method according to claim 4, wherein, according to the second signal amplitude ratio, the first frequency domain near-end audio signal and the second frequency domain near-end audio signal, the The second frequency-domain near-end audio signal is subjected to nonlinear echo cancellation to obtain a frequency-domain near-end audio enhancement signal, including:

Acquiring the product of the second signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal as a reference amplitude;

Acquiring the minimum value of the reference amplitude and the amplitude of the second frequency-domain near-end audio signal as the amplitude of the frequency-domain near-end audio enhancement signal;

A frequency-domain near-end audio enhancement signal is obtained according to the amplitude of the frequency-domain near-end audio enhancement signal and the phase of the first frequency-domain near-end audio signal.
The audio signal processing method according to claim 4 or 5, wherein the second signal amplitude ratio is obtained by the following formula:

Mask(n,k)=min{1-RCr(n,k),1-RY p r(n,k)};

Wherein, Mask(n,k) is the second signal amplitude ratio, RCr(n,k) is the cross-correlation coefficient between the frequency-domain far-end reference audio signal and the frequency-domain near-end acquisition audio signal, RYp r(n,k) is the cross-correlation coefficient between the frequency domain far-end reference audio signal and the second frequency domain near-end audio signal, n is the number of frame sequences, k is the number of center frequency sequences, 0<n≤N , 0<k≤K, N is the total number of frames, K is the total number of frequency bands.
An audio signal processing device, comprising:

The signal acquisition unit is configured to: acquire a near-end audio signal, a far-end reference audio signal, and a first near-end audio signal obtained by performing linear echo cancellation on the near-end audio signal;

The frequency-domain conversion unit is configured to: perform time-frequency conversion on the near-end audio signal, the far-end reference audio signal, and the first near-end audio signal, respectively, to obtain the frequency-domain near-end audio signal, frequency A domain far-end reference audio signal and a first frequency-domain near-end audio signal;

The deep noise reduction unit is configured to: perform deep learning and noise reduction on the amplitude of the first frequency-domain near-end audio signal to obtain a second frequency-domain near-end audio signal;

A non-linear elimination unit configured to: based on the frequency-domain far-end reference audio signal, the frequency-domain near-end collected audio signal, the first frequency-domain near-end audio signal, and the second frequency-domain near-end audio signal, performing nonlinear echo cancellation on the second frequency-domain near-end audio signal to obtain a frequency-domain near-end audio enhancement signal;

The time-domain transformation unit is configured to: perform time-frequency inverse transformation on the frequency-domain near-end audio enhancement signal to obtain the near-end audio enhancement signal.
The audio signal processing device according to claim 7, wherein the deep noise reduction unit is configured to:

Through the trained noise reduction neural network model, the magnitude of the near-end audio signal in the first frequency domain is subjected to deep learning and noise reduction to obtain the magnitude of the near-end audio signal in the second frequency domain;

The second frequency-domain near-end audio signal is obtained according to the amplitude of the second frequency-domain near-end audio signal and the phase of the first frequency-domain near-end audio signal.
The audio signal processing device according to claim 8, wherein the deep noise reduction unit is configured to:

Inputting the magnitude of the near-end audio signal in the first frequency domain into the trained noise reduction neural network model to obtain a first signal magnitude ratio, wherein the first signal magnitude ratio is equal to the second frequency domain near-end audio signal A predicted value of the ratio of the amplitude of the end audio signal to the amplitude of the first frequency domain near end audio signal;

According to the first signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal, the amplitude of the second frequency-domain near-end audio signal is obtained, wherein the amplitude of the second frequency-domain near-end audio signal is the product of the first signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal.
The audio signal processing device according to claim 7, wherein the non-linear cancellation unit is configured to:

Correlating the frequency-domain far-end reference audio signal with the frequency-domain near-end acquisition audio signal and the second-frequency near-end audio signal in each frequency band to obtain a second signal amplitude ratio of each frequency band;

According to the second signal amplitude ratio, the first frequency-domain near-end audio signal, and the second frequency-domain near-end audio signal, nonlinear echo cancellation is performed on the second frequency-domain near-end audio signal to obtain a frequency Domain near-end audio enhancement signal.
The audio signal processing device according to claim 10, wherein the non-linear cancellation unit is configured to:

Acquiring the product of the second signal amplitude ratio and the amplitude of the first frequency-domain near-end audio signal as a reference amplitude;

Acquiring the minimum value of the reference amplitude and the amplitude of the second frequency-domain near-end audio signal as the amplitude of the frequency-domain near-end audio enhancement signal;

A frequency-domain near-end audio enhancement signal is obtained according to the amplitude of the frequency-domain near-end audio enhancement signal and the phase of the first frequency-domain near-end audio signal.
The audio signal processing device according to claim 10 or 11, wherein the second signal amplitude ratio is obtained by the following formula:

Mask(n,k)=min{1-RCr(n,k),1-RY p r(n,k)};

Wherein, Mask(n,k) is the second signal amplitude ratio, RCr(n,k) is the cross-correlation coefficient between the frequency-domain far-end reference audio signal and the frequency-domain near-end acquisition audio signal, RYp r(n,k) is the cross-correlation coefficient between the frequency domain far-end reference audio signal and the second frequency domain near-end audio signal, n is the number of frame sequences, k is the number of center frequency sequences, 0<n≤N , 0<k≤K, N is the total number of frames, K is the total number of frequency bands.
An electronic device comprising:

at least one processor;

at least one memory storing computer-executable instructions,

Wherein, the computer-executable instructions, when executed by the at least one processor, cause the at least one processor to perform the following steps:

Acquiring a near-end audio signal, a far-end reference audio signal, and a first near-end audio signal obtained by performing linear echo cancellation on the near-end audio signal;

performing time-frequency transformation on the near-end collected audio signal, the far-end reference audio signal and the first near-end audio signal respectively, to obtain the frequency-domain near-end collected audio signal, the frequency-domain far-end reference audio signal and the first frequency domain near-end audio signal;

performing deep learning noise reduction on the amplitude of the first frequency-domain near-end audio signal to obtain a second frequency-domain near-end audio signal;

Based on the frequency-domain far-end reference audio signal, the frequency-domain near-end collected audio signal, the first frequency-domain near-end audio signal, and the second frequency-domain near-end audio signal, the second frequency domain Non-linear echo cancellation is performed on the near-end audio signal to obtain a near-end audio enhancement signal in the frequency domain;

Inverse time-frequency transform is performed on the near-end audio enhancement signal in the frequency domain to obtain the near-end audio enhancement signal.
A non-volatile computer-readable storage medium, wherein, when the instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is caused to perform the following steps:

Acquiring a near-end audio signal, a far-end reference audio signal, and a first near-end audio signal obtained by performing linear echo cancellation on the near-end audio signal;

performing time-frequency transformation on the near-end collected audio signal, the far-end reference audio signal and the first near-end audio signal respectively, to obtain the frequency-domain near-end collected audio signal, the frequency-domain far-end reference audio signal and the first frequency domain near-end audio signal;

performing deep learning noise reduction on the amplitude of the first frequency-domain near-end audio signal to obtain a second frequency-domain near-end audio signal;

Based on the frequency-domain far-end reference audio signal, the frequency-domain near-end collected audio signal, the first frequency-domain near-end audio signal, and the second frequency-domain near-end audio signal, the second frequency domain Non-linear echo cancellation is performed on the near-end audio signal to obtain a near-end audio enhancement signal in the frequency domain;

Inverse time-frequency transform is performed on the near-end audio enhancement signal in the frequency domain to obtain the near-end audio enhancement signal.
A computer program product comprising computer instructions, wherein said computer instructions, when executed by at least one processor, implement the following steps:

Acquiring a near-end audio signal, a far-end reference audio signal, and a first near-end audio signal obtained by performing linear echo cancellation on the near-end audio signal;

performing time-frequency transformation on the near-end collected audio signal, the far-end reference audio signal and the first near-end audio signal respectively, to obtain the frequency-domain near-end collected audio signal, the frequency-domain far-end reference audio signal and the first frequency domain near-end audio signal;

performing deep learning noise reduction on the amplitude of the first frequency-domain near-end audio signal to obtain a second frequency-domain near-end audio signal;

Based on the frequency-domain far-end reference audio signal, the frequency-domain near-end collected audio signal, the first frequency-domain near-end audio signal, and the second frequency-domain near-end audio signal, the second frequency domain Non-linear echo cancellation is performed on the near-end audio signal to obtain a near-end audio enhancement signal in the frequency domain;

Inverse time-frequency transform is performed on the near-end audio enhancement signal in the frequency domain to obtain the near-end audio enhancement signal.