Method and electronic device for reducing echo residue

Abstract

The present application discloses a method and an electronic device for reducing echo residue. The method for reducing echo residue can be applied to the electronic device, and comprises: performing echo cancellation on a voice input signal according to an echo reference signal to obtain an echo cancellation signal; converting the echo reference signal into a reference spectrum signal of each frame by fast Fourier transform; converting the echo cancellation signal into a voice spectrum signal of each frame by fast Fourier transform; obtaining an a priori signal-to-noise ratio of a current frame by using the reference spectrum signal of the current frame and the voice spectrum signal of the current frame according to an additive noise principle; filtering the voice spectrum signal of the current frame by a Wiener filter coefficient of the current frame determined by the a priori signal-to-noise ratio of the current frame to obtain a target spectrum signal of each frame; and converting the target spectrum signal of each frame by inverse fast Fourier transform to obtain a target voice signal. Therefore, the method for reducing echo residue can accurately filter out residual echo.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of voice echo cancellation, and in particular to a method for reducing residual echo and an electronic device. BACKGROUND

[0002] With the rapid development of wireless communication technology and intelligent speech recognition technology, people have higher and higher requirements for voice quality. However, the existence of echo will affect the voice quality, resulting in poor user experience or affecting the speech recognition rate.

[0003] For echo cancellation, the method of Acoustic Echo Canceller (AEC) is generally used at present. The echo path characteristics are estimated by an adaptive filter to generate a signal (hereinafter referred to as an echo signal) that is the same as the echo, and then the echo signal is subtracted from the signal received by the sound receiving module to achieve the effect of echo cancellation. However, in actual application, due to various factors, such as background noise, nonlinear distortion of the sound receiving module and the loudspeaker, loudspeaker time delay, echo path mutation, etc., the echo cannot be completely cancelled by the adaptive filter, and a small amount of residual echo will exist, thereby affecting the user's auditory experience or the speech recognition rate.

[0004] Therefore, how to provide a method for reducing residual echo, which can accurately filter out residual echo, is a problem that those skilled in the art are eager to solve. SUMMARY

[0005] The embodiments of the present application provide a method for reducing residual echo and an electronic device, which can effectively solve the problem of residual echo affecting the user's auditory experience or the speech recognition rate in the prior art.

[0006] In order to solve the above technical problems, the present application is implemented as follows:

[0007] In a first aspect, a method for reducing residual echo is provided, which includes the following steps: performing echo cancellation processing on a voice input signal according to an echo reference signal to obtain an echo cancellation signal; converting the echo reference signal into a reference spectrum signal of each frame by fast Fourier transform; converting the echo cancellation signal into a voice spectrum signal of each frame by fast Fourier transform; obtaining a prior signal-to-noise ratio of a current frame by using the reference spectrum signal of the current frame and the voice spectrum signal of the current frame according to the additive noise principle; filtering the voice spectrum signal of the current frame by using the Wiener filter coefficient of the current frame determined by the prior signal-to-noise ratio of the current frame to obtain a target spectrum signal of each frame; and converting the target spectrum signal of each frame by inverse fast Fourier transform to obtain a target voice signal.

[0008] In a second aspect, an electronic device is provided, which includes a non-transitory program storage medium and one or more processors. The non-transitory program storage medium is configured to store program code. The one or more processors are configured to execute the program code to perform: performing echo cancellation processing on a voice input signal according to an echo reference signal to obtain an echo cancelled signal; converting the echo reference signal into a reference spectral signal of each frame by a fast Fourier transform; converting the echo cancelled signal into a voice spectral signal of each frame by a fast Fourier transform; obtaining a prior signal-to-noise ratio of a current frame according to an additive noise principle using the reference spectral signal of the current frame and the voice spectral signal of the current frame; filtering the voice spectral signal of the current frame by a Wiener filter coefficient of the current frame determined by the prior signal-to-noise ratio of the current frame to obtain a target spectral signal of each frame; and converting the target spectral signal of each frame by an inverse fast Fourier transform to obtain a target voice signal.

[0009] In the embodiments of the present application, the method for reducing echo residue can estimate the echo residue directly using the echo reference signal and obtain the Wiener filter coefficient through the prior and posterior signal-to-noise ratios, so that the Wiener filter coefficient can be obtained by a low complexity method, the residual echo can be accurately filtered out, the pollution of the residual echo to the voice can be reduced, the good voice quality can be ensured, and the method for reducing echo residue can be conveniently applied to electronic devices such as embedded products. In addition, when the method for reducing echo residue is applied to the field of intelligent voice recognition, the voice recognition rate can be improved. Furthermore, when the method for reducing echo residue is applied to communication such as network conference call, the call quality can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0010] The accompanying drawings, which are included to provide a further understanding of the present application and are incorporated in and constitute a part of this application, illustrate embodiments of the present application and serve to explain the present application. In the drawings:

[0011] Figure 1 An embodiment block diagram of an electronic device according to the present application;

[0012] Figure 2 An embodiment method flowchart of a method for reducing echo residue according to the present application;

[0013] Figure 3 Another embodiment method flowchart of a method for reducing echo residue according to the present application;

[0014] Figure 4 Still another embodiment method flowchart of a method for reducing echo residue according to the present application; and

[0015] Figure 5This is a flowchart of another embodiment of the method for reducing echo residue according to this application. Detailed Implementation

[0016] The embodiments of the present invention will be described below with reference to the accompanying drawings. In these drawings, the same reference numerals denote the same or similar components or method flows.

[0017] It must be understood that the use of terms such as "comprising" or "including" in this specification is intended to indicate the presence of specific technical features, values, method steps, work processes, components and / or components, but does not preclude the addition of more technical features, values, method steps, work processes, components, or any combination thereof.

[0018] It is important to understand that when a component is described as "connected" or "coupled" to another component, it can be a direct connection or coupling to other components, and there may be intermediate components. Conversely, when a component is described as "directly connected" or "directly coupled" to another component, there are no intermediate components.

[0019] Please see Figure 1 This is a block diagram of an embodiment of an electronic device according to this application. Figure 1 As shown, the electronic device 100 includes a non-transitory program storage medium 110 and a processor 120, wherein the non-transitory program storage medium 110 is used to store program code; and the processor 120 is connected to the non-transitory program storage medium 110 and is used to execute the program code.

[0020] In this embodiment, the number of processors 120 may be, but is not limited to, one; for example, the number of processors 120 may be multiple. The processor 120 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The non-transitory program storage medium 110 may be a read-only memory (ROM), a magnetic disk, or an optical disk, etc.

[0021] In the embodiment, the processor 120 executes the program code stored in the non-transitory program storage medium 110 to perform: performing echo cancellation on the voice input signal according to the echo reference signal to obtain an echo cancelled signal; converting the echo reference signal into a reference spectral signal of each frame by fast Fourier transform; converting the echo cancelled signal into a voice spectral signal of each frame by fast Fourier transform; obtaining a prior signal-to-noise ratio of a current frame by using the reference spectral signal of the current frame and the voice spectral signal of the current frame according to the additive noise principle; filtering the voice spectral signal of the current frame by a Wiener filter coefficient of the current frame determined by the prior signal-to-noise ratio of the current frame to obtain a target spectral signal of each frame; and converting the target spectral signal of each frame by inverse fast Fourier transform to obtain a target voice signal.

[0022] More specifically, the processor 120 performs fast Fourier transform on the echo cancelled signal and the echo reference signal in units of frames after removing the echo from the voice input signal by the echo reference signal, converts the echo cancelled signal and the echo reference signal from time domain to frequency domain, and obtains a reference spectral signal N[k] of each frame and a voice spectral signal Y[k] of each frame, and further obtains an amplitude value N mag [k] corresponding to the reference spectral signal N[k] in the frequency domain and an amplitude value Y mag [k] corresponding to the voice spectral signal Y[k] in the frequency domain, where 0≤k≤M-1, M is a positive integer greater than 1, k is a frequency point index, Y mag [k] = magnitude(Y[k]), and N mag [k] = magnitude(N[k]).

[0023] In an embodiment, the processor 120 simplifies the prior signal-to-noise ratio of each frame to a ratio of the target signal of the same frame to the amplitude value of the corresponding reference spectral signal according to the minimum mean square error (i.e. S mag [k] is the amplitude value corresponding to the target signal (i.e. the residual echo removed signal) in the frequency domain, and i is a frame index). That is, the prior signal-to-noise ratio of the current frame is the ratio of the non-reference spectral signal part in the voice spectral signal of the current frame to the reference spectral signal of the current frame. More specifically, the echo reference signal (e.g. music sound) is regarded as noise by the processor 120, so that the echo cancelled signal obtained by the processor 120 performing echo cancellation on the voice input signal according to the echo reference signal can include human voice (e.g. voice command) and other noise (e.g. wind noise, environmental sound), and therefore the processor 120 can simplify the prior signal-to-noise ratio of each frame to the ratio of the residual echo removed signal of the same frame to the amplitude value corresponding to the reference spectral signal in the frequency domain according to the minimum mean square error.

[0024] In an embodiment, the processor 120 simplifies the post SNR of each frame according to the minimum mean square error as a ratio of the amplitude value of the speech spectrum signal of the same frame to the amplitude value of the corresponding reference spectrum signal (i.e. That is, the post SNR of the current frame is the ratio of the speech spectrum signal of the current frame to the reference spectrum signal of the current frame.

[0025] In an embodiment, the processor 120 adds the amplitude value of the target signal (i.e. the echo residual signal removed) corresponding to the speech spectrum signal of each frame to the amplitude value of the reference spectrum signal according to the additive noise principle to obtain the amplitude value of the speech spectrum signal of each frame (i.e. mag [k] = S mag [k] + N mag [k]); then, the processor 120 obtains the relationship among the prior SNR of the current frame, the prior SNR of the previous frame and the post SNR of the current frame according to the post SNR of each frame simplified by the minimum mean square error and the prior SNR of each frame and the additive noise principle: ζ i (k) = ζ i-1 (k) + (1 - a)(y i (k) - 1), where ζ i (k) is the prior SNR, y i (k) is the post SNR, i is the frame index, and a is a smoothing parameter with a value ranging from 0 to 1. When a is close to 0, the prior SNR result is close to the prior SNR of the current frame obtained according to the maximum likelihood criterion, and when a is close to 1, the prior SNR result is close to the prior SNR of the previous frame, so the smoothing parameter is a balance parameter for the two parts. That is, the processor 120 can obtain the prior SNR of the current frame by smoothing the post SNR of the current frame and the prior SNR of the previous frame based on the additive noise principle.

[0026] In an embodiment, the processor 120 can obtain the Wiener filter coefficient of the current frame by the prior SNR of the current frame, i.e. H(ω) is the Wiener filter coefficient of the current frame, and ω is the frequency point index; the processor 120 can multiply the speech spectrum signal Y(ω) of the current frame and the Wiener filter coefficient H(ω) of the current frame to perform filtering, and obtain the target spectrum signal of the current frame, and then obtain the target spectrum signal of each frame based on the above method.

[0027] In the embodiment, the processor 120 can perform inverse fast Fourier transform on the target spectrum signal of each frame, convert the target spectrum signal of each frame from the frequency domain to the time domain, and integrate the conversion result into the target speech signal (i.e. the signal without echo residual).

[0028] Therefore, the electronic device 100 can obtain the Wiener filtering coefficient by a low complexity method, and accurately filter out the residual echo, reduce the pollution of the residual echo to the speech, and ensure good speech quality. In addition, since the method for obtaining the Wiener filtering coefficient by the processor 120 has low complexity, the electronic device 100 can be an embedded product in actual implementation.

[0029] In an embodiment, the electronic device 100 can further include a power amplifier circuit 130, a speaker 140, and a back sampling circuit 150. The power amplifier circuit 130 is connected to the processor 120, the speaker 140 is connected to the power amplifier circuit 130, and the back sampling circuit 150 is connected to the power amplifier circuit 130. The power amplifier circuit 130 is configured to output an audio analog signal, the speaker 140 is configured to play the audio analog signal, and the back sampling circuit 150 is configured to collect the audio analog signal and perform analog-digital conversion on the audio analog signal to obtain the echo reference signal. Since the audio analog signal collected by the back sampling circuit 150 connected to the power amplifier circuit 130 is closest to the audio analog signal played by the speaker 140, the echo reference signal obtained based on the audio analog signal collected by the back sampling circuit 150 can reduce noise in the echo reference signal and improve accuracy.

[0030] It should be noted that the power amplifier circuit 130 and the back sampling circuit 150 can be implemented by pure hardware or hardware combined with firmware and / or software. The firmware or software can be stored on a machine-readable storage medium, such as a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a register, a magnetic disk, an optical disk, a hard drive, or any other suitable storage medium, or any combination thereof, and can be executed by one or more general-purpose or special-purpose programmable microprocessors.

[0031] In an embodiment, the electronic device 100 can further include a memory 160 connected to the back sampling circuit 150 and the processor 120, configured to store the echo reference signal. In an embodiment, the memory 160 and the processor 120 can be integrated into a single chip system (System on a Chip, SoC). In another embodiment, the memory 160 can include a high-speed random access memory, and can further include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory device.

[0032] In an embodiment, the electronic device 100 can further comprise a sound receiving module 170 connected to the processor 120, for receiving a voice receiving signal; the processor 120 is further configured to perform a first pre-processing procedure on the voice receiving signal to obtain the voice input signal. The first pre-processing procedure comprises beamforming and de-noising processing on the voice receiving signal to obtain the voice input signal. In an embodiment, the sound receiving module 170 can be, but is not limited to, a microphone.

[0033] In another embodiment, the sound receiving module 170 is configured to receive the voice input signal, and the processor 120 is configured to perform a second pre-processing procedure on the target voice signal to obtain a voice output signal after obtaining the target voice signal. The second pre-processing procedure comprises beamforming and de-noising processing on the target voice signal.

[0034] Please refer to Figure 1 and Figure 2 , Figure 2 is an embodiment method flowchart of the method for reducing echo residue according to the present application. In the embodiment, the method for reducing echo residue can be applied to the electronic device 100 and comprises the following steps: performing echo cancellation processing on a voice input signal according to an echo reference signal to obtain an echo cancellation signal (step 210); converting the echo reference signal into a reference spectral signal of each frame by fast Fourier transform (step 220); converting the echo cancellation signal into a voice spectral signal of each frame by fast Fourier transform (step 230); obtaining a prior signal-to-noise ratio of a current frame by using the reference spectral signal of the current frame and the voice spectral signal of the current frame according to an additive noise principle (step 240); filtering the voice spectral signal of the current frame by using a Wiener filter coefficient of the current frame determined by the prior signal-to-noise ratio of the current frame to obtain a target spectral signal of each frame (step 250); and converting the target spectral signal of each frame by inverse fast Fourier transform to obtain a target voice signal (step 260). Steps 210 to 260 can be performed by the processor 120, and detailed descriptions can be referred to the related descriptions of the electronic device 100 above, which will not be repeated here.

[0035] In an embodiment, step 240 can comprise obtaining the prior signal-to-noise ratio of the current frame by using a smoothing of the prior signal-to-noise ratio of the current frame and the prior signal-to-noise ratio of a previous frame based on the additive noise principle. Detailed descriptions can be referred to the related descriptions of the electronic device 100 above, which will not be repeated here.

[0036] In an embodiment, the prior signal-to-noise ratio of the current frame is a ratio of a non-reference spectral signal part in the voice spectral signal of the current frame to the reference spectral signal of the current frame. Detailed descriptions can be referred to the related descriptions of the electronic device 100 above, which will not be repeated here.

[0037] In an embodiment, the post-signal-to-noise ratio of the current frame is a ratio of the speech spectrum signal of the current frame and the reference spectrum signal of the current frame. For details, please refer to the relevant description of the electronic device 100 above, which will not be repeated here.

[0038] In an embodiment, please refer to Figure 1 and Figure 3 , Figure 3 is another embodiment method flowchart of the method for reducing echo residue according to the present application. The method for reducing echo residue can further include, before step 210: collecting an audio analog signal output by the power amplifier circuit 130 of the electronic device 100, and performing analog-to-digital conversion on the audio analog signal to obtain the echo reference signal (step 310). Wherein, the audio analog signal is played through the loudspeaker 140 of the electronic device 100; step 310 can be performed by the echo collecting circuit 150, and for details, please refer to the relevant description of the electronic device 100 above, which will not be repeated here. It should be noted that, in order to avoid Figure 3 the figure is too dense, Figure 3 steps 220 to 260 are omitted.

[0039] In an embodiment, please refer to Figure 1 and Figure 4 , Figure 4 is still another embodiment method flowchart of the method for reducing echo residue according to the present application. The method for reducing echo residue can further include, before step 210: obtaining a speech receiving signal received by the receiving module 170 of the electronic device 100 (step 410); and performing a first preprocessing procedure on the speech receiving signal to obtain the speech input signal (step 420). Wherein, steps 410 and 420 can be performed by the processor 120, and for details, please refer to the relevant description of the electronic device 100 above, which will not be repeated here. It should be noted that, in order to avoid Figure 4 the figure is too dense, Figure 4 steps 220 to 260 are omitted.

[0040] In an embodiment, the first preprocessing procedure of step 420 includes: performing beamforming and denoising processing on the speech receiving signal.

[0041] In an embodiment, please refer to Figure 1 and Figure 5 , Figure 5For another embodiment of the method for reducing echo residual according to the present application, a method flow chart is shown. The method for reducing echo residual can further comprise, after step 260, performing a second pre-processing procedure on the target speech signal to obtain a speech output signal (step 510). Step 510 can be performed by the processor 120, and the detailed description can refer to the related description of the electronic device 100 above, which will not be repeated here. It should be noted that, in order to avoid Figure 5 the figure surface is too dense, Figure 5 the drawing steps 210 to 250 are omitted.

[0042] In an embodiment, the second pre-processing procedure of step 510 comprises performing beamforming and denoising processing on the target speech signal.

[0043] In summary, the method for reducing echo residual according to the embodiments of the present application can obtain the Wiener filter coefficient by directly using the echo reference signal to estimate the echo residual, and by using the prior and posterior signal-to-noise ratios, so that the Wiener filter coefficient can be obtained by a low-complexity method, and the residual echo can be accurately filtered out, the pollution of the residual echo to the speech is reduced, the good speech quality is ensured, and the method for reducing echo residual is convenient to apply to electronic devices such as embedded products. In addition, when the method for reducing echo residual is applied to the field of intelligent speech recognition, the speech recognition rate can be improved. In addition, when the method for reducing echo residual is applied to communication such as network conference call, the call quality can be improved.

[0044] Although the above-described components are included in the drawings of the present application, it does not exclude the use of more other additional components without violating the spirit of the invention, so as to achieve better technical effects.

[0045] Although the present application is described using the above embodiments, it should be noted that these descriptions are not intended to limit the present application. On the contrary, the present application encompasses modifications and similar arrangements that are obvious to those skilled in the art. Therefore, the scope of the claims should be interpreted in the broadest way to include all obvious modifications and similar arrangements.

Claims

1. A method of reducing echo residual, applied to embedded products, characterized in that, The method comprises the following steps: performing echo cancellation on the voice input signal according to the echo reference signal to obtain an echo cancellation signal; converting the echo reference signal into a reference spectrum signal of each frame through fast Fourier transform; converting the echo cancellation signal into a voice spectrum signal of each frame through the fast Fourier transform; obtaining a prior signal-to-noise ratio of a current frame according to an additive noise principle by using the reference spectrum signal of the current frame and the voice spectrum signal of the current frame; filtering the voice spectrum signal of the current frame by using the Wiener filter coefficient of the current frame calculated through the prior signal-to-noise ratio of the current frame to obtain a target spectrum signal of each frame; and converting the target spectrum signal of each frame through inverse fast Fourier transform to obtain a target voice signal. The method further comprises the following steps:

2. The method of claim 1, wherein, obtaining a voice receiving signal received by a receiving module of an electronic device; and performing a first preprocessing procedure on the voice receiving signal to obtain the voice input signal. The first preprocessing procedure comprises performing beamforming and denoising processing on the voice receiving signal. The method further comprises the following steps:

3. The method of claim 2, wherein, collecting an audio analog signal output by a power amplifier circuit of the electronic device, and performing analog-to-digital conversion on the audio analog signal to obtain the echo reference signal, wherein the audio analog signal is played through a loudspeaker of the electronic device.

4. The method of claim 1, wherein, The prior signal-to-noise ratio of the current frame is a ratio of a non-reference spectrum signal part in the voice spectrum signal of the current frame to the reference spectrum signal of the current frame. The step of obtaining the prior signal-to-noise ratio of the current frame according to the additive noise principle by using the reference spectrum signal of the current frame and the voice spectrum signal of the current frame comprises the following step: obtaining the prior signal-to-noise ratio of the current frame based on smoothing of a posterior signal-to-noise ratio of the current frame and a prior signal-to-noise ratio of a previous frame according to the additive noise principle.

5. The method of claim 1, wherein, The posterior signal-to-noise ratio of the current frame is a ratio of the voice spectrum signal of the current frame to the reference spectrum signal of the current frame.

6. The method of reducing echo residue of claim 1, wherein, The method further comprises the following steps:

7. The method of reducing echo residue as claimed in claim 6, wherein, performing a second preprocessing procedure on the target voice signal to obtain a voice output signal.

8. The method of reducing echo residue of claim 1, wherein, The second preprocessing procedure comprises the following steps: performing beamforming and denoising processing on the target voice signal.

9. The method of reducing echo residue as claimed in claim 8, wherein, The method comprises the following steps: a non-transitory program storage medium for storing program code; and 10. An electronic device of an embedded product, characterized by, one or more processors for executing the program code to perform the following steps: performing echo cancellation on the voice input signal according to the echo reference signal to obtain an echo cancellation signal; converting the echo reference signal into a reference spectrum signal of each frame through fast Fourier transform; converting the echo cancellation signal into a voice spectrum signal of each frame through the fast Fourier transform; obtaining a prior signal-to-noise ratio of a current frame according to an additive noise principle by using the reference spectrum signal of the current frame and the voice spectrum signal of the current frame; filtering the voice spectrum signal of the current frame by using the Wiener filter coefficient of the current frame calculated through the prior signal-to-noise ratio of the current frame to obtain a target spectrum signal of each frame; and converting the target spectrum signal of each frame through inverse fast Fourier transform to obtain a target voice signal. The electronic device of the embedded product further comprises the following steps: ​ ​ ​ 11. The electronic device of claim 10, wherein, ​ The power amplifier circuit is connected to the one or more processors and configured to output an audio analog signal; The speaker is connected to the power amplifier circuit and configured to play the audio analog signal; and The echo recovery circuit is connected to the power amplifier circuit and configured to collect the audio analog signal and perform analog-digital conversion on the audio analog signal to obtain the echo reference signal.

12. The electronic device of claim 11, wherein, The electronic device of the embedded product further comprises: The memory is connected to the echo recovery circuit and the one or more processors and configured to store the echo reference signal.

13. The electronic device of claim 10, wherein, The one or more processors are further configured to obtain the prior signal-to-noise ratio of the current frame by smoothing the posterior signal-to-noise ratio of the current frame and the prior signal-to-noise ratio of the previous frame based on the additive noise principle.

14. The electronic device of claim 13, wherein, The posterior signal-to-noise ratio of the current frame is the ratio of the speech spectrum signal of the current frame to the reference spectrum signal of the current frame.

15. The electronic device of claim 10, wherein, The electronic device of the embedded product further comprises a radio module configured to receive a speech receiving signal; and the one or more processors are further configured to perform a first preprocessing procedure on the speech receiving signal to obtain the speech input signal.

16. The electronic device of claim 15, wherein, The one or more processors are further configured to perform beamforming and noise reduction processing on the speech receiving signal to obtain the speech input signal.

17. The electronic device of Claim 10, wherein, The prior signal-to-noise ratio of the current frame is the ratio of the non-reference spectrum signal part in the speech spectrum signal of the current frame to the reference spectrum signal of the current frame.

18. The electronic device of Claim 10, wherein, The one or more processors are further configured to perform a second preprocessing procedure on the target speech signal to obtain a speech output signal.

19. The electronic device of claim 18, wherein, The one or more processors are further configured to perform beamforming and noise reduction processing on the target speech signal to obtain the speech output signal.

Images