WO2018188282A1 - 回声消除方法、装置、会议平板及计算机存储介质 - Google Patents

回声消除方法、装置、会议平板及计算机存储介质 Download PDF

Info

Publication number
WO2018188282A1
WO2018188282A1 PCT/CN2017/104391 CN2017104391W WO2018188282A1 WO 2018188282 A1 WO2018188282 A1 WO 2018188282A1 CN 2017104391 W CN2017104391 W CN 2017104391W WO 2018188282 A1 WO2018188282 A1 WO 2018188282A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
adaptive filter
reference signal
frequency domain
correlation
Prior art date
Application number
PCT/CN2017/104391
Other languages
English (en)
French (fr)
Inventor
刘荣
程雪峰
Original Assignee
广州视源电子科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 广州视源电子科技股份有限公司 filed Critical 广州视源电子科技股份有限公司
Publication of WO2018188282A1 publication Critical patent/WO2018188282A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04M—TELEPHONIC COMMUNICATION
    • H04M9/00—Arrangements for interconnection not involving centralised switching
    • H04M9/08—Two-way loud-speaking telephone systems with means for conditioning the signal, e.g. for suppressing echoes for one or both directions of traffic
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise
    • G10L21/0232—Processing in the frequency domain
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0264—Noise filtering characterised by the type of parameter measurement, e.g. correlation techniques, zero crossing techniques or predictive techniques
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/21—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being power information
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L2021/02082—Noise filtering the noise being echo, reverberation of the speech
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise
    • G10L2021/02161—Number of inputs available containing the signal or the noise to be suppressed
    • G10L2021/02163—Only one microphone

Definitions

  • the present application relates to the field of voice processing technologies, and in particular, to an echo cancellation method, apparatus, conference tablet, and computer storage medium.
  • FIG. 1A it is a schematic diagram of a voice interaction scenario in the related art.
  • User A and user B perform a conference call.
  • the voice of user A is transmitted by the electronic device A1 to the user B side through the network after being collected by the microphone A2.
  • the electronic device B1, the electronic device B1 plays the voice of the user A through the speaker B3.
  • the user B is also talking.
  • the microphone B2 collects the voice signal, not only the user B's voice is collected, but also the user A voice played by the speaker B3 at this time is collected.
  • the voice signal collected by the microphone B2 (the superimposed result of the user B's voice and the user A's voice played by the speaker B3) is not transmitted to the user A side without the echo cancellation processing, and the user A will hear the speaker A3 playing the user B.
  • the voice of the voice and the voice of the user A's own voice, this phenomenon is the echo phenomenon.
  • echo cancellation is required in the related art, that is, the voice collected by the microphone B2 is eliminated, and the user A speaking voice (ie, the echo signal) played by the speaker B3 is eliminated.
  • the voice signal collected by the microphone is: the superposition result of the user B's speech sound and the echo signal, how to accurately determine the echo signal is a key step affecting the echo cancellation effect.
  • FIG. 1B it is a schematic diagram of signal propagation in the related art.
  • the voice played and played in the air will be interfered by many environmental factors until it is collected and processed by the microphone. Therefore, after the original signal passes through the above propagation process (ie, the echo path), The echo signal collected with the microphone will be quite different.
  • an adaptive filter technique is adopted, and original voice data (also commonly referred to as a far-end signal and a reference signal) is used as a reference, and the original voice data is estimated to be collected by the microphone after being subjected to an echo path such as play and space propagation.
  • the echo signal obtained at the time.
  • the adaptive filter can automatically take a specific algorithm (such as a minimum mean square error algorithm or a recursive least squares algorithm) based on the estimation of the statistical characteristics of the input signal (original speech data) and the output signal (predicted echo signal).
  • the filter coefficients are calculated, and the calculation result is used as an output signal by calculating the correlation between the filter coefficients and the input signal sequence.
  • the present invention provides an echo cancellation method, apparatus, conference tablet and computer storage medium to solve the related art how to enable an adaptive filter to quickly and accurately determine a suitable adaptive filter coefficient for a complex voice interaction environment.
  • an echo cancellation method comprising:
  • the adaptive filter coefficients are updated according to a correlation between the residual signal and the reference signal.
  • the updating the adaptive filter coefficients according to the correlation between the residual signal and the reference signal includes:
  • the calculating, by using a correlation coefficient, a correlation between the residual signal and the reference signal includes:
  • a correlation coefficient between a power spectrum of the residual signal in the frequency domain and a power spectrum of the reference signal in the frequency domain is calculated.
  • the calculating a correlation coefficient between a power spectrum of the residual signal in a frequency domain and a power spectrum of the reference signal in a frequency domain includes:
  • the correlation coefficient is calculated by the following formula:
  • cohxe is the correlation coefficient
  • xPow(f) is a power spectrum of the reference signal in a frequency domain
  • ePow(f) is a power spectrum of the residual signal in the frequency domain
  • the xePow(f) is a correlation power spectrum of the xPow(f) and the conjugate signal of the residual signal.
  • the estimating the echo signal corresponding to the reference signal by using the adaptive filter coefficient includes:
  • the correlation coefficient includes: a correlation coefficient between the residual signal and each frequency point of the reference signal in a frequency domain;
  • Determining, according to the correlation coefficient, a step factor for adjusting the adaptive filter coefficient including one or more of the following manners:
  • Correlating coefficients corresponding to the respective frequency points in the frequency domain are respectively used as step factors of the respective frequency points of the adaptive filter coefficients in the frequency domain;
  • the updating the adaptive filter coefficients according to the step size factor includes:
  • the adaptive filter coefficients are updated by the following formula:
  • W k (f) is the adaptive filter coefficient for the frequency point f at the kth time
  • ⁇ (f) is the step size factor of the frequency point f
  • X(k) is the frequency domain reference
  • the signal, E(f), is a frequency domain residual signal of the residual signal in the frequency domain.
  • an echo cancellation apparatus comprising:
  • a signal acquisition module configured to: obtain a reference signal input to the speaker for playing, and acquire an acquisition signal of the microphone;
  • An echo cancellation module configured to: estimate an echo signal corresponding to the reference signal by using an adaptive filter coefficient, and Eliminating the echo signal in the acquired signal, obtaining a residual signal and outputting;
  • a coefficient updating module configured to: update the adaptive filter coefficients according to a correlation between the residual signal and the reference signal.
  • a conference tablet includes an echo cancellation device, and the echo cancellation device is configured to:
  • the adaptive filter coefficients are updated according to a correlation between the residual signal and the reference signal.
  • a computer storage medium where the program medium includes program instructions, where the program instructions include:
  • the adaptive filter coefficients are updated according to a correlation between the residual signal and the reference signal.
  • the filter in consideration of the far-end speech, if the current echo cancellation effect of the adaptive filter is better, the filter will eliminate the echo signal in the acquired signal of the microphone, that is, the residual signal is not mixed. Too many remote users' sounds; if the current echo cancellation effect of the adaptive filter is poor, the residual signal will be mixed with some remote users. Based on this, the embodiment can determine whether the echo in the residual signal is cleaned by determining the correlation between the residual signal and the reference signal, and determine the echo cancellation effect of the adaptive filter.
  • the adaptive filter coefficient can be updated in time to make the adaptive filter perform calculation under the adjusted coefficient, and enhance Echo cancellation effect; if the correlation between the residual signal and the reference signal is low, it means that the current echo cancellation effect of the adaptive filter is better, then the adaptive filter can be stabilized under the current coefficient to maintain the current echo cancellation. effect.
  • the solution of the embodiment of the present application only needs to increase the update of the adaptive filter coefficients on the basis of the traditional algorithm, so the added cost is low.
  • FIG. 1A is a schematic diagram of a scenario of a voice interaction in the related art.
  • Fig. 1B is a schematic diagram of signal propagation in the related art.
  • FIG. 2A is a flowchart of an echo cancellation method according to an exemplary embodiment of the present application.
  • FIG. 2B is a schematic diagram of an echo canceller according to an exemplary embodiment of the present application.
  • FIG. 2C is a schematic diagram of an application scenario of an echo cancellation method according to an exemplary embodiment of the present application.
  • FIG. 3 is a block diagram of an electronic device in which an echo cancellation device is located according to an exemplary embodiment of the present application.
  • FIG. 4 is a block diagram of an echo canceling apparatus according to an exemplary embodiment of the present application.
  • first, second, third, etc. may be used to describe various information in this application, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other.
  • first information may also be referred to as the second information without departing from the scope of the present application.
  • second information may also be referred to as the first information.
  • word "if” as used herein may be interpreted as "when” or “when” or “in response to a determination.”
  • the echo message scheme of the embodiment of the present application can be applied to a voice interaction device in a voice interaction scenario, such as a conference call, an in-vehicle system, an IP phone, and a human-machine interaction.
  • the voice interaction device can be a cellular phone, a media player, an audio device, or the like.
  • Conference tablet devices, in-vehicle devices, telephones, gaming devices, tablet computers, notebook computers, desktop computers or televisions, and the like require electronic devices that involve voice processing and have certain computing power.
  • the present application provides an echo cancellation method according to an exemplary embodiment, which can be applied to an electronic device, and the method can include the following steps 201 to 203:
  • step 201 a reference signal input to the speaker for playing is acquired, and an acquisition signal of the microphone is acquired.
  • step 202 an echo signal corresponding to the reference signal is estimated by using an adaptive filter coefficient, and the echo signal is cancelled from the acquired signal to obtain a residual signal and output.
  • step 203 the adaptive filter coefficients are updated according to a correlation between the residual signal and the reference signal.
  • the microphone and speaker are involved in the voice interaction scenario.
  • the microphone may be a microphone
  • the speaker may be a speaker
  • the microphone and speaker may be independent of each other and connected to a computer.
  • the microphone and speaker can also be two devices that are configured in the same electronic device, such as smartphones and tablets, with built-in microphones and speakers.
  • the electronic device is generally configured with an echo canceller, as shown in FIG. 2B, which is a schematic diagram of an echo canceller according to an exemplary embodiment of the present application.
  • the solution of the embodiment of the present application can be applied to an echo canceller to Echo cancellation is performed as necessary.
  • the input data of the echo canceller includes a microphone acquisition signal (Mic) and a reference signal (Ref) played through the speaker, and the electronic device can acquire Mic and Ref by reading the data of the hardware buffer area.
  • Mic microphone acquisition signal
  • Ref reference signal
  • the adaptive echo technique can be used to simulate the echo path, that is, the corresponding echo signal is estimated based on Ref, and the echo signal is eliminated in the Mic to obtain an output residual signal.
  • the operation process of the adaptive filter is based on the estimation of the statistical characteristics of the input signal and the output signal, and adopts a specific algorithm to automatically adjust the filter coefficients to achieve an optimal filtering characteristic.
  • the adaptive filter can update and adjust the filter coefficients for each sample of the input signal sequence according to a specific algorithm (such as a minimum mean square error algorithm or a recursive least squares algorithm), and pass the filter coefficients and the input signal sequence. The correlation calculation between the two finally results in a residual signal.
  • the convergence phase In the operation of the adaptive filter, the convergence phase is involved (that is, the filter starts from an initial state, and according to the set rules, the filter coefficients are adjusted according to the observed input signal and output signal, so that the filter is continuously approaching The process of the optimal coefficient), the convergence phase requires the convergence of the adaptive filter to start very fast. For example, when the local user and the remote user start voice interaction, the adaptive filter starts to learn quickly. The optimal effect is that the remote user does not have time to talk, or the remote user starts to talk, and the adaptive filter is It has already converged; after convergence, the far-end user starts to talk, the echo signal starts to generate, and the adaptive filter coefficient needs to be stable, that is, the adaptive filter needs to be stable. The echo cancels the state.
  • the echo path may be changed. Once the change occurs, the adaptive filter can be judged because the adaptive filter needs to relearn the adaptive filter coefficient. Adaptive filters are required to be kept up to date to ensure that the changing echo path can be tracked.
  • the adaptive filter usually solves the adaptive filter coefficients in real time by using an algorithm such as a minimum mean square error algorithm or a recursive least squares algorithm.
  • an algorithm such as a minimum mean square error algorithm or a recursive least squares algorithm.
  • the algorithm uses the steepest descent method to estimate the coefficient vector of the next moment by iteratively estimating the current filter coefficient vector from the gradient of the mean square error.
  • an embodiment of the present application proposes a scheme for determining how to perform adaptive filter coefficient update by determining whether the echo in the residual signal is cleared.
  • the filter will eliminate the echo signal in the acquired signal of the microphone, that is, the residual signal will not be mixed with too many remote users. Sound; if the current echo cancellation effect of the adaptive filter is poor, the residual signal will be mixed with the sound of some remote users.
  • the embodiment can first determine whether the echo in the residual signal is cleared. Specifically, whether the echo of the residual signal is cleared can be determined by determining the correlation between the residual signal and the reference signal. If the correlation between the residual signal and the reference signal is high, it means that there are many echoes in the residual signal, and the current echo cancellation effect of the adaptive filter is poor, and the adaptive filter coefficient can be updated in time to make the adaptive filter adjust. The operation is performed under the coefficient to enhance the echo cancellation effect; if the correlation between the residual signal and the reference signal is low, it means that there is no more echo in the residual signal, and the current echo cancellation effect of the adaptive filter is better. The adaptive filter can be stabilized under the current coefficients to maintain the current echo cancellation effect. The solution of the embodiment of the present application only needs to increase the update of the adaptive filter coefficients on the basis of the traditional algorithm, so the added cost is low.
  • the reference signal input to the speaker for playback is usually a time domain signal x(k), which can be converted into a frequency domain signal X(f) by a Fourier algorithm:
  • X(f) FFT[x(k-M),...,x(k),...,x(k+M-1)], taking the first M elements;
  • the FFT is a Fast Fourier Transformatio, that is, a fast Fourier transform
  • k represents the time
  • f represents the frequency point
  • M represents the length of the adaptive filter.
  • the signal In the frequency domain, the signal consists of multiple frequency points, so when estimating the echo signal, it can be:
  • Determining a frequency domain reference signal of the reference signal in a frequency domain Determining a frequency domain reference signal of the reference signal in a frequency domain, where the frequency domain reference signal includes a plurality of frequency points.
  • W(f) represents the adaptive filter coefficient and Y(f) represents the estimated echo signal.
  • the estimated frequency domain echo signal Y(f) can be transformed into a time domain signal y(f):
  • e(k) represents the residual signal and d(k) represents the acquired signal of the microphone.
  • This step completes the calculation of the time domain residual signal.
  • the residual signal can be sent to the remote user.
  • the embodiment of the present application may determine the echo cancellation effect of the current time adaptive filter to determine the adaptive filter coefficient at the next moment according to the current echo effect.
  • the echo cancellation effect may be determined according to the correlation between the residual signal and the reference signal.
  • the correlation between the residual signal and the reference signal is correspondingly adjusted for the adaptive filter coefficient, and can be flexibly configured in practical applications.
  • the experimental results of different environments or different devices may be used to determine the correlation.
  • the relative relationship between the height and the coefficient adjustment range, and the like, is not limited in this embodiment.
  • the updating the adaptive filter coefficients according to the correlation between the residual signal and the reference signal includes:
  • a correlation coefficient is calculated for indicating a correlation between the residual signal and the reference signal.
  • the correlation between the residual signal and the reference signal that is, the degree of similarity between the residual signal and the reference signal.
  • waveform comparison, power spectrum comparison, phase spectrum comparison or spectrum can be used to analyze the two.
  • the correlation coefficient is not limited in this embodiment. Through the size of the correlation coefficient, the corresponding step factor can be determined according to actual needs, and the step factor can be used to quickly adjust or decrease the adaptive filter coefficient.
  • the power spectrum of the residual signal in the frequency domain may be calculated and the power spectrum of the reference signal in the frequency domain is correlated.
  • Coefficient for two signals, By correlating the correlation between the power spectra of the signals in the frequency domain indicating the correlation between the two signals, a relatively accurate correlation coefficient value can be obtained, and the amount of calculation is small.
  • the calculating a correlation coefficient between a power spectrum of the residual signal in a frequency domain and a power spectrum of the reference signal in a frequency domain including:
  • the correlation coefficient is calculated by the following formula:
  • cohxe is the correlation coefficient
  • the xPow(f) is a power spectrum of the reference signal in a frequency domain
  • the ePow(f) is a power spectrum of the residual signal in a frequency domain
  • the xePow (f) is a correlation power spectrum of the conjugate signal of the xPow(f) and the residual signal.
  • the residual signal can be converted to the frequency domain E(f):
  • the correlation coefficient is calculated based on the power spectrum of the current time signal in the frequency domain.
  • the signal is composed of a plurality of frequency points, and the echo cancellation is performed in the frequency domain, the processing speed is faster, and the echo cancellation effect is better, so the correlation coefficient may include: the residual signal and the reference signal are Correlation coefficient corresponding to each frequency point in the frequency domain.
  • the first type the correlation coefficient corresponding to each frequency point in the frequency domain is used as a step factor of each frequency point of the adaptive filter coefficient in the frequency domain; Multiple frequency points are formed, so the calculated number of adaptive step factors is equal to the number of frequency points. In this way, each frequency point has a dedicated step factor for adjusting the adaptive filter coefficients, so the coefficients The adjustment is finer, the echo cancellation is cleaner, and the error is smaller.
  • the second method is to calculate an average value of correlation coefficients corresponding to each frequency point in the frequency domain, and use the average value as a step factor of each frequency point of the adaptive filter coefficient in the frequency domain; In this way, all the frequency points adopt the average step factor, and the original data relationship can still be maintained between the processed frequency points, and the processing speed is fast.
  • the adaptive filter may adopt a Normalized Least Mean Square (NLMS) algorithm. Therefore, based on the foregoing algorithm, updating the adaptive filter coefficients according to the step factor may be:
  • the adaptive filter coefficients are updated by the following formula:
  • W k (f) is the adaptive filter coefficient for the frequency point f at the kth time
  • ⁇ (f) is the step size factor of the frequency point f
  • X(k) is the frequency domain reference
  • the signal, E(f), is a frequency domain residual signal of the residual signal in the frequency domain.
  • the frequency domain residual signal may refer to a normalized frequency domain residual signal, and the frequency domain residual signal is normalized for each frequency point, so that the filter divergence problem in the double-talk state can be solved.
  • the adaptive filter needs to be updated at this time, and the correlation between the reference signal and the residual signal is strong, but since the residual signal contains the local voice, if the adaptive filter coefficient is updated according to the residual signal, There may be a problem of filter divergence. Therefore, after the residual signal is normalized, the adaptive filter coefficient is updated to solve the divergence problem.
  • the solution of the embodiment of the present application deliberately determines the adaptive step size factor by the correlation between the residual signal and the reference signal, and can perform maximum echo cancellation for each call state.
  • the embodiment of the present application is described again in conjunction with a schematic diagram of an application scenario shown in FIG. 2C.
  • the 2C is related to two users, and the electronic device in FIG. 2C is described by taking a conference tablet as an example.
  • the conference tablet is used for real-time conference call between the two users, and the speaker and the microphone are integrated in the conference tablet.
  • the conference tablet can apply the solution of the embodiment of the present application to perform echo cancellation during the call.
  • the electronic device can also be a smart phone, a personal computer or a tablet, and the like.
  • the call state between users usually includes four types:
  • the first type, double-ended silence that is, the local user and the remote user have no sound.
  • the reference signal transmitted by the remote user is weak, because the local user does not make a sound, so
  • the residual signal is also weak.
  • both the reference signal and the residual signal are weak, usually the reference signal and the residual signal are white noise. Since the reference signal and the residual signal are collected in different environments, the correlation between the two is usually small, so the application of the present application is applied.
  • the step factor for adjusting the adaptive filter is also small. In the actual situation, no echo is generated in the double-ended silent state, and the adaptive filter does not need to perform more echo cancellation work.
  • the second type is the remote end: the local user does not make a sound, and the remote user sends a sound.
  • the reference signal transmitted by the remote user is stronger, and the local user does not make a sound.
  • the local user's microphone mainly collects the echo signal.
  • the residual signal output by the adaptive filter may still contain more echoes, so Both the reference signal and the residual signal are highly correlated, thus enabling adaptive filter coefficients to be adjusted faster.
  • the adaptive filter In the adaptive filter stabilization phase, the adaptive filter can better eliminate the echo, so the residual signal will not be mixed with more echo signals. Therefore, the correlation between the reference signal and the residual signal is small, so the adjustment to the adaptive filter is also small. In the actual situation, since the adaptive filter has entered the stable phase, better echo cancellation can be performed. At this time, it is not necessary to greatly adjust the adaptive filter coefficients, and only the filter coefficients need to be stabilized.
  • the update processing of the adaptive filter coefficients by the application embodiment can meet the actual requirements.
  • the third type the near-end single-speaking: that is, the local user makes a sound, and the remote user does not emit a sound.
  • the reference signal transmitted by the remote user is almost zero, and the local user's microphone is mainly collected.
  • the local user's voice does not collect more echo signals, so the residual signal strength is stronger.
  • the reference signal is almost zero, and the residual signal is strong, so the correlation coefficient is almost zero, and the step factor is almost zero.
  • the fourth type, double-ended speech that is, the local user makes a sound, and the remote user also makes a sound.
  • the adaptive filter needs to update the coefficients in time. Since the remote user makes a sound, the reference signal is strong; and the local user also makes a sound, so the residual signal strength is also strong. Since there are two user-supplied sounds in the reference signal and the residual signal, the adjustment effect on the adaptive filter may not be obvious in the first three cases, but the situation of double-end simultaneous speech may be less, this embodiment The scheme can still meet the adaptive filter coefficient adjustment for most cases.
  • the present application also provides an embodiment of an echo cancellation device, a conference tablet, and a computer storage medium.
  • Embodiments of the echo cancellation device of the present application can be applied to an electronic device.
  • the device embodiment may be implemented by software, or may be implemented by hardware or a combination of hardware and software.
  • the processor of the electronic device in which the computer is located reads the corresponding computer program instructions in the non-volatile memory into the memory.
  • FIG. 3 a hardware structure diagram of an electronic device in which the echo cancellation device is located, except for the processor 310, the memory 330, the network interface 320, and the non-volatile device shown in FIG.
  • the electronic device in which the device 331 is located in the embodiment may also include other hardware according to the actual function of the electronic device, and details are not described herein again.
  • FIG. 4 is a block diagram of an echo canceling apparatus according to an exemplary embodiment of the present application, the apparatus includes:
  • the signal acquisition module 41 is configured to: acquire a reference signal input to the speaker for playing, and acquire an acquisition signal of the microphone.
  • the echo cancellation module 42 is configured to: estimate an echo signal corresponding to the reference signal by using an adaptive filter coefficient, and cancel the echo signal from the acquired signal to obtain a residual signal and output.
  • the coefficient update module 43 is configured to: update the adaptive filter coefficients according to a correlation between the residual signal and the reference signal.
  • the coefficient update module 43 is further configured to:
  • the coefficient update module 43 is further configured to:
  • a correlation coefficient between a power spectrum of the residual signal in the frequency domain and a power spectrum of the reference signal in the frequency domain is calculated.
  • the coefficient update module 43 is further configured to:
  • the correlation coefficient is calculated by the following formula:
  • cohxe is the correlation coefficient
  • the xPow(f) is a power spectrum of the reference signal in a frequency domain
  • the ePow(f) is a power spectrum of the residual signal in a frequency domain
  • the xePow (f) is a correlation power spectrum of the conjugate signal of the xPow(f) and the residual signal.
  • the echo cancellation module 42 includes:
  • the correlation coefficient includes: a correlation coefficient between the residual signal and each frequency point of the reference signal in a frequency domain;
  • the coefficient update module 43 is further configured to:
  • a step factor for adjusting the adaptive filter coefficients is determined according to one or more of the following:
  • Correlating coefficients corresponding to the respective frequency points in the frequency domain are respectively used as step factors of the respective frequency points of the adaptive filter coefficients in the frequency domain;
  • the coefficient update module 43 is further configured to:
  • the adaptive filter coefficients are updated by the following formula:
  • Wk(f) is the adaptive filter coefficient for the frequency point f at the kth time
  • ⁇ (f) is the step size factor of the frequency point f
  • X(k) is the frequency domain reference signal E(f) is a frequency domain residual signal of the residual signal in the frequency domain.
  • the device embodiment since it basically corresponds to the method embodiment, reference may be made to the partial description of the method embodiment.
  • the device embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separate, and the components displayed as modules may or may not be physical modules, ie may be located A place, or it can be distributed to multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the objectives of the present application. Those of ordinary skill in the art can understand and implement without any creative effort.
  • the embodiment of the present application further provides a conference tablet, where the conference tablet includes an echo cancellation device, and the echo cancellation device is configured to:
  • the adaptive filter coefficients are updated according to a correlation between the residual signal and the reference signal.
  • the embodiment of the present application is a computer storage medium, where the storage medium stores program instructions, where the program instructions include:
  • the adaptive filter coefficients are updated according to a correlation between the residual signal and the reference signal.
  • the application can take the form of a computer program product embodied on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) in which program code is embodied.
  • Computer-usable storage media includes both permanent and non-persistent, removable and non-removable media, and information storage can be implemented by any method or technology.
  • the information can be computer readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory.
  • PRAM phase change memory
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • RAM random access memory
  • ROM read only memory
  • EEPROM electrically erasable programmable read only memory
  • flash memory or other memory technology
  • CD-ROM compact disc
  • DVD digital versatile disc
  • magnetic cassette magnetic tape storage or other magnetic storage

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Human Computer Interaction (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Quality & Reliability (AREA)
  • Telephone Function (AREA)
  • Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
  • Circuit For Audible Band Transducer (AREA)

Abstract

关于回声消除方法、装置、会议平板及计算机存储介质，方法包括：获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号(201)；获取自适应滤波器系数，利用自适应滤波器系数估计参考信号对应的回声信号；从采集信号中消除回声信号，获得残留信号并输出(202)；根据残留信号与参考信号之间相关性的高低更新自适应滤波器系数(203)。开创性地通过残留信号与参考信号之间的相关性来对自适应滤波器系数进行调整，能够进行极大限度的回声消除，并且只需在传统的算法基础上，增加对自适应滤波器系数的更新，新增成本较低。

Description

回声消除方法、装置、会议平板及计算机存储介质 技术领域
本申请涉及语音处理技术领域，尤其涉及回声消除方法、装置、会议平板及计算机存储介质。
背景技术
在电话会议系统、车载系统、IP电话、人机交互等涉及语音交互的场景中，通常会出现回声现象。如图1A所示，是相关技术中一种语音交互的场景示意图，用户A和用户B进行电话会议，用户A说话的声音，经过麦克风A2采集后，由电子设备A1通过网络传播至用户B侧的电子设备B1，电子设备B1通过音箱B3播放用户A说话声音。此时，用户B也正在说话，麦克风B2在采集语音信号时，不仅会采集到用户B说话声音，也会采集到此时音箱B3播放的用户A说话声音。
假设麦克风B2所采集的语音信号(用户B说话声音和音箱B3播放的用户A说话声音的叠加结果)没有经过回声消除处理，就传输至用户A侧，用户A会听到音箱A3播放出用户B说话声音以及用户A自己的说话声音，这种现象即回声现象。
针对上述现象，相关技术中需要进行回声消除，也即是将麦克风B2所采集的语音，消除掉音箱B3播放的用户A说话声音(即回声信号)。在消除回声过程中，由于麦克风采集的语音信号是：用户B说话声音和回声信号的叠加结果，如何准确地确定回声信号，是影响回声消除效果的关键步骤。
如图1B所示，是相关技术中一种信号传播示意图。输入至音箱的原始语音数据经过音箱处理后进行播放、播放的语音在空气中传播又会受到较多环境因素干扰、直至经过麦克风采集处理，因此原始信号经过上述传播过程(即回音路径)后，与麦克风采集到的回声信号将会有较大差异。相关技术中采用自适应滤波器技术，以原始语音数据(通常也称为远端信号、参考信号)作为参考，预估出该原始语音数据在经过播放、空间传播等回音路径后，至麦克风采集时所得到的回声信号。
自适应滤波器能以输入信号(原始语音数据)和输出信号(预估的回声信号)的统计特性的估计为依据，采取特定算法(如最小均方误差算法或递归最小平方算法等)自动地计算滤波器系数，通过滤波器系数与输入信号序列之间的相关计算，将计算结果作为输出信号。
实际应用中，麦克风的采集语音信号时受环境影响的干扰较大，如何使自适应滤波器能针对复杂的语音交互环境，快速精确地确定合适的自适应滤波器系数，是亟待解决的技术问题。
发明内容
基于此，本发明提供了回声消除方法、装置、会议平板及计算机存储介质，以解决相关技术中如何使自适应滤波器能针对复杂的语音交互环境、快速精确地确定合适的自适应滤波器系数的技术问题。
根据本申请实施例的第一方面，提供一种回声消除方法，所述方法包括：
获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
在一个可选的实现方式中，所述根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数，包括：
计算用于指示所述残留信号与所述参考信号之间相关性的相关系数；
根据所述相关系数确定用于调整所述自适应滤波器系数的步长因子，利用所述步长因子更新所述自适应滤波器系数。
在一个可选的实现方式中，所述计算用于指示所述残留信号与所述参考信号之间相关性的相关系数，包括：
计算所述残留信号在频域上的功率谱与所述参考信号在频域上的功率谱的相关系数。
在一个可选的实现方式中，所述计算所述残留信号在频域上的功率谱与所述参考信号在频域上的功率谱的相关系数，包括：
通过如下公式计算所述相关系数：
Figure PCTCN2017104391-appb-000001
其中，cohxe为所述相关系数，所述xPow(f)为所述参考信号在频域上的功率谱，所述 ePow(f)为所述残留信号在频域上的功率谱，所述xePow(f)为所述xPow(f)与所述残留信号的共轭信号的相关功率谱。
在一个可选的实现方式中，所述利用所述自适应滤波器系数估计所述参考信号对应的回声信号，包括：
确定所述参考信号在频域上的频域参考信号，所述频域参考信号中包括多个频点；
利用所述自适应滤波器系数计算所述频域参考信号中的每个频点对应的估计频点，获得所述回声信号。
在一个可选的实现方式中，所述相关系数包括：所述残留信号与所述参考信号在频域上的各个频点对应的相关系数；
所述根据所述相关系数确定用于调整所述自适应滤波器系数的步长因子，包括如下一种或多种方式：
将所述在频域上的各个频点对应的相关系数分别作为所述自适应滤波器系数在频域上各个频点的步长因子；
统计所述在频域上的各个频点对应的相关系数的平均值，将所述平均值作为所述自适应滤波器系数在频域上每个频点的步长因子；
确定所述在频域上的各个频点对应的相关系数的中位数，将所述中位数作为所述自适应滤波器系数在频域上每个频点的步长因子。
在一个可选的实现方式中，所述根据所述步长因子更新所述自适应滤波器系数，包括：
通过如下公式更新所述自适应滤波器系数：
Figure PCTCN2017104391-appb-000002
其中，f为频点，Wk(f)为第k时刻针对频点f的自适应滤波器系数，μ(f)为频点f的步长因子，X(k)为所述频域参考信号，E(f)为所述残留信号在频域上的频域残留信号。
根据本申请实施例的第二方面，提供一种回声消除装置，所述装置包括：
信号获取模块，用于：获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
回声消除模块，用于：利用自适应滤波器系数估计所述参考信号对应的回声信号，并从 所述采集信号中消除所述回声信号，获得残留信号并输出；
系数更新模块，用于：根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
根据本申请实施例的第三方面，提供一种会议平板，所述会议平板包括回声消除装置，所述回声消除装置用于：
获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
根据本申请实施例的第四方面，提供一种计算机存储介质，所述存储介质中存储有程序指令，所述程序指令包括：
获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
本申请的实施例提供的技术方案可以包括以下有益效果：
本申请实施例中，考虑到远端讲话的情况下，若自适应滤波器当前的回声消除效果较好，滤波器会将麦克风的采集信号中的回声信号消除干净，即残留信号中不会夹杂太多远端用户的声音；若自适应滤波器当前的回声消除效果较差，则残留信号中会夹杂一些远端用户的声音。基于此，本实施例可以通过判断残留信号与参考信号的相关性，来确定残留信号中回声是否清除干净，确定自适应滤波器的回声消除效果。若残留信号与参考信号的相关性较高，则表示自适应滤波器当前的回声消除效果较差，可以及时更新自适应滤波器系数，使自适应滤波器在调整后的系数下进行运算，增强回声消除效果；若残留信号与参考信号的相关性较低，则表示自适应滤波器当前的回声消除效果较好，则可以使自适应滤波器稳定在当前系数下进行运算，保持当前的回声消除效果。本申请实施例方案只需在传统的算法基础上，增加对自适应滤波器系数的更新，因此新增成本较低。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，并不能限制本申请。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本申请的实施例，并与说明书一起用于解释本申请的原理。
图1A是相关技术中一种语音交互的场景示意图。
图1B是相关技术中一种信号传播示意图。
图2A是本申请根据一示例性实施例示出的一种回声消除方法的流程图。
图2B是本申请根据一示例性实施例示出的一种回声消除器的示意图。
图2C是本申请根据一示例性实施例示出的一种回声消除方法的应用场景示意图。
图3是本申请根据一示例性实施例示出的一种回声消除装置所在电子设备的框图。
图4是本申请根据一示例性实施例示出的一种回声消除装置的框图。
具体实施方式
这里将详细地对示例性实施例进行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本申请相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本申请的一些方面相一致的装置和方法的例子。
在本申请使用的术语是仅仅出于描述特定实施例的目的，而非旨在限制本申请。在本申请和所附权利要求书中所使用的单数形式的“一种”、“所述”和“该”也旨在包括多数形式，除非上下文清楚地表示其他含义。还应当理解，本文中使用的术语“和/或”是指并包含一个或多个相关联的列出项目的任何或所有可能组合。
应当理解，尽管在本申请可能采用术语第一、第二、第三等来描述各种信息，但这些信息不应限于这些术语。这些术语仅用来将同一类型的信息彼此区分开。例如，在不脱离本申请范围的情况下，第一信息也可以被称为第二信息，类似地，第二信息也可以被称为第一信息。取决于语境，如在此所使用的词语“如果”可以被解释成为“在……时”或“当……时”或“响应于确定”。
本申请实施例的回声消息方案可应用于电话会议、车载系统、IP电话、人机交互等涉及语音交互场景中的语音交互设备，语音交互设备可以是蜂窝电话、媒体播放器、音响设备、 会议平板设备、车载设备、电话机、游戏设备、平板计算机、笔记本计算机、台式计算机或电视机等等需要涉及语音处理且具有一定计算能力的电子设备。
如图2A所示，是本申请根据一示例性实施例提供的一种回声消除方法，可应用于电子设备，该方法可包括如下步骤201至203：
在步骤201中，获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号。
在步骤202中，利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出。
在步骤203中，根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
语音交互场景中涉及麦克风和扬声器，在某些例子中，如图1A的例子中，麦克风可以是话筒，扬声器可以是音箱，麦克风和扬声器可以是相互独立的、与计算机连接的设备。在另一些例子中，麦克风和扬声器还可以是配置在同一个电子设备的两个器件，如智能手机和平板电脑等电子设备内置有麦克风和扬声器。
电子设备通常配置有回声消除器，如图2B所示，是本申请根据一示例性实施例示出的一种回声消除器的示意图，本申请实施例的方案可应用于回声消除器中，以在必要时进行回声消除。回声消除器的输入数据包括麦克风的采集信号(Mic)和通过扬声器播放的参考信号(Ref)，电子设备可以通过读取硬件缓存区的数据而获取到Mic和Ref。
本实施例中，可以采用自适应滤波器技术模拟回声路径，即基于Ref估计出对应的回声信号，并在Mic中消除掉回声信号，得到可输出的残留信号。自适应滤波器的运算过程，是以输入信号和输出信号的统计特性的估计为依据，采取特定算法自动地调整滤波器系数，使其达到最佳滤波特性的一种算法。自适应滤波器可以对输入信号序列的每一个样值，按特定的算法(如最小均方误差算法或递归最小平方算法等)，更新、调整滤波器系数，通过滤波器系数与输入信号序列之间的相关计算，最终获得残留信号。
自适应滤波器的运算过程中，涉及收敛阶段(也即是滤波器从某个初始状态出发，按照设定的规则，依据观测到的输入信号和输出信号，调整滤波器系数，使其不断逼近最优系数的过程)，收敛阶段要求自适应滤波器开始运转后收敛要非常快。举例来说，当本端用户与远端用户开始进行语音交互，自适应滤波器开始快速学习，最优的效果是远端用户还来不及说话，或者是远端用户开始说话，自适应滤波器就已经收敛好了；收敛好之后，远端用户开始说话，回声信号开始产生，自适应滤波器系数需要稳定，也即是自适应滤波器需处于稳定 的回声消除状态。由于麦克风的采集语音信号时受环境影响的干扰较大，回音路径可能是变化的，一旦出现变化，自适应滤波器要能判断出来，因为自适应滤波器需要重新学习自适应滤波器系数，因此需要自适应滤波器随时保持更新状态，以保证能够追踪变化的回音路径。
相关技术中，自适应滤波器通常采用最小均方误差算法或递归最小平方算法等算法实时求解自适应滤波器系数。以最小均方误差算法为例，该算法利用最陡下降法，由均方误差的梯度估计现时刻滤波器系数向量迭代计算下一个时刻的系数向量。
本申请实施例针对如何快速精确地确定合适的自适应滤波器系数的问题，提出了一种通过判断残留信号中回声是否清除干净，来确定如何进行自适应滤波器系数更新的方案。
考虑到远端讲话的情况下，若自适应滤波器当前的回声消除效果较好，滤波器会将麦克风的采集信号中的回声信号消除干净，即残留信号中不会夹杂太多远端用户的声音；若自适应滤波器当前的回声消除效果较差，则残留信号中会夹杂一些远端用户的声音。
基于此，本实施例可以先判断残留信号中回声是否清除干净，具体的，可以通过判断残留信号与参考信号相关性的高低来确定残留信号中回声是否清除干净。若残留信号与参考信号的相关性较高，则表示残留信号中夹杂较多回声，自适应滤波器当前的回声消除效果较差，可以及时更新自适应滤波器系数，使自适应滤波器在调整后的系数下进行运算，增强回声消除效果；若残留信号与参考信号的相关性较低，则表示表示残留信号中不会夹杂较多回声，自适应滤波器当前的回声消除效果较好，则可以使自适应滤波器稳定在当前系数下进行运算，保持当前的回声消除效果。本申请实施例方案只需在传统的算法基础上，增加对自适应滤波器系数的更新，因此新增成本较低。
其中，对于步骤202，输入至扬声器进行播放的参考信号通常为时域信号x(k)，可以通过傅里叶算法转换为频域信号X(f)：
X(f)＝FFT[x(k-M),...,x(k),...,x(k+M-1)]，取前M个元素；
FFT为Fast Fourier Transformatio，即快速傅里叶变换；
k表示时刻，f表示频点，M表示自适应滤波器的长度。
在频域上，信号由多个频点构成，因此，在估计回声信号时，可以是：
确定所述参考信号在频域上的频域参考信号，所述频域参考信号中包括多个频点。
利用所述自适应滤波器系数计算所述频域参考信号中的每个频点对应的估计频点，获得所述回声信号。
上述处理过程可以由如下公式表示：
Figure PCTCN2017104391-appb-000003
其中，W(f)表示自适应滤波器系数，Y(f)表示估计的回声信号。
接着，可以将估计的频域回声信号Y(f)变换为时域信号y(f)：
y(k)＝IFFT[Y(f)]
计算在时域上的残留信号：
e(k)＝d(k)-y(k)
其中，e(k)表示残留信号，d(k)表示麦克风的采集信号。
此步骤完成对时域残留信号的计算，作为自适应滤波器的输出信号，残留信号可以发送给远端用户。
对于当前时刻所计算的残留信号，本申请实施例可以确定当前时刻自适应滤波器的回声消除效果，以根据当前回声效果确定下一时刻的自适应滤波器系数。具体的，回声消除效果可以根据所述残留信号与所述参考信号之间相关性的高低而确定。而残留信号与参考信号的相关性的高低，对应地对于自适应滤波器系数的调整幅度，在实际应用中可以灵活配置，例如可以是通过对不同环境或不同设备等等的实验结果，确定相关性高低与系数调整幅度的相对关系等等，本实施例对此不作限定。
在一个可选的实现方式中，所述根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数，包括：
计算用于指示所述残留信号与所述参考信号之间相关性的相关系数。
根据所述相关系数确定用于调整所述自适应滤波器系数的步长因子，利用所述步长因子更新所述自适应滤波器系数。
残留信号与参考信号之间相关性，也即是残留信号与参考信号之间的相似程度，在实际应用中，可以采用波形比较、功率谱对比、相谱对比或频谱等多种方式分析两者的相关系数，本实施例对此不作限定。通过相关系数的大小，可以根据实际需要确定相应的步长因子，利用步长因子可以快速地对自适应滤波器系数进行增大或减小的调整。
其中，对于残留信号与参考信号之间相关性，在一个可选的实现方式中，可以是计算所述残留信号在频域上的功率谱与所述参考信号在频域上的功率谱的相关系数，对于两个信号， 通过信号在频域上的功率谱之间相关性高低指示两个信号的相关性高低，可以获得较为精确的相关系数值，且运算量较小。
具体的，所述计算所述残留信号在频域上的功率谱与所述参考信号在频域上的功率谱的相关系数，包括是：
通过如下公式计算所述相关系数：
Figure PCTCN2017104391-appb-000004
其中，cohxe为所述相关系数，所述xPow(f)为所述参考信号在频域上的功率谱，所述ePow(f)为所述残留信号在频域上的功率谱，所述xePow(f)为所述xPow(f)与所述残留信号的共轭信号的相关功率谱。
举例来说，可以将残留信号转换为频域E(f)：
E(f)＝FFT[0M个0,e(k)]
计算参考信号X(f)在频域上的功率谱xPow(f)：
xPow(f)＝||X(f)||2
计算残留信号E(f)在频域上的功率谱ePow(f)：
ePow(f)＝||E(f)||2
计算频域参考信号X(f)与共轭残留信号E*(f)的相关功率谱xePow(f)：
xePow(f)＝||X(f)·E*(f)||2
计算参考信号X(f)与共轭残留信号E*(f)的相关系数cohxe：
Figure PCTCN2017104391-appb-000005
由上述计算过程可知，相关系数是根据当前时刻信号在频域上的功率谱而计算得到。在频域上，信号由多个频点构成，而在频域内进行回声消除，其处理速度较快，且回声消除效果较好，因此相关系数可以包括：所述残留信号与所述参考信号在频域上的各个频点对应的相关系数。
所述根据所述相关系数确定用于调整所述自适应滤波器系数的步长因子，包括如下一种 或多种方式：
第一种、将所述在频域上的各个频点对应的相关系数分别作为所述自适应滤波器系数在频域上各个频点的步长因子；由于当前时刻内所采集的一段信号由多个频点构成，因此计算得到的自适应步长因子个数等于频点个数，此种方式下，每个频点都有专属的步长因子用于调整自适应滤波器系数，因此系数调整更为精细，回声消除更为干净，误差更小。
第二种、统计所述在频域上的各个频点对应的相关系数的平均值，将所述平均值作为所述自适应滤波器系数在频域上每个频点的步长因子；此种方式下，所有频点采用平均值作为的步长因子，处理后频点之间仍能保持原有的数据关系，且处理速度较快。
第三种、确定所述在频域上的各个频点对应的相关系数的中位数，将所述中位数作为所述自适应滤波器系数在频域上每个频点的步长因子。此种方式下，由于中位数的取值能消除一段信号中的误差，能体现了信号相关性的集中趋势，自适应滤波的效果更好。
对于计算得到的步长因子，可用于对自适应滤波器系数的更新，更新后的自适应滤波器将用于下一时刻的自适应滤波处理。本实施例中，自适应滤波器可以采用最小均方误差(Normalized Least Mean Square，NLMS)算法，因此，基于上述算法，根据所述步长因子更新所述自适应滤波器系数，可以是：
通过如下公式更新所述自适应滤波器系数：
Figure PCTCN2017104391-appb-000006
其中，f为频点，Wk(f)为第k时刻针对频点f的自适应滤波器系数，μ(f)为频点f的步长因子，X(k)为所述频域参考信号，E(f)为所述残留信号在频域上的频域残留信号。
实际应用中，上述频域残留信号可以是指归一化后的频域残留信号，将频域残留信号针对各个频点进行归一化处理，可以解决双端讲话状态下的滤波器发散问题。双端讲话状态下，此时自适应滤波器需要更新，而参考信号与残留信号的相关性又强，但由于残留信号中包含有本端话音，若根据该残留信号更新自适应滤波器系数，有可能会造成滤波器发散问题，因而将残留信号进行归一化处理后，再进行自适应滤波器系数更新，即可解决发散问题。
本申请实施例方案开创性地通过残留信号与参考信号之间的相关性确定自适应步长因子，对各通话状态均可进行极大限度的回声消除。结合图2C所示的一种应用场景示意图，对本申请实施例再次进行说明。
图2C中涉及两个用户，图2C中的电子设备以会议平板为例进行说明，两个用户之间都采用会议平板进行实时会议通话，会议平板中集成有扬声器和麦克风。会议平板可以应用本申请实施例的方案，以在通话过程中进行回音消除。可以理解，在其他例子中，电子设备还可以是智能手机、个人计算机或平板电脑等等。
在上述场景中，用户之间的通话状态通常包括四种：
第一种、双端静默：即本端用户和远端用户都没发出声音的情况，对于本端来说，远端用户所传输的参考信号强度较弱，由于本端用户没有发出声音，因此残留信号也较弱。参考信号和残留信号都较弱的情况下，通常参考信号和残留信号都是白噪声，由于参考信号和残留信号是在不同环境采集的，两者的相关性通常较小，因此应用本申请实施例方案时，针对自适应滤波器进行调整的步长因子也较小。而实际情况中，双端静默状态下不会产生回声，自适应滤波器确实不需进行较多的回声消除工作。
第二种、远端单讲：即本端用户没有发出声音，远端用户发出声音的情况，对于本端来说，远端用户所传输的参考信号较强，而由于本端用户没有发出声音，本端用户的麦克风主要采集到的是回声信号。
其中，在自适应滤波器初始阶段，由于自适应滤波器暂未达到收敛(也即是回声消除能力较弱阶段)，此时自适应滤波器输出的残留信号中可能仍夹杂较多回声，因此参考信号和残留信号两者相关性较大，因此能使自适应滤波器系数更快调整。而实际情况中，在自适应滤波器初始阶段，确实需要对自适应滤波器系数进行较快调整，以使自适应滤波器系数更快逼近最优系数。
到自适应滤波器稳定阶段，此时自适应滤波器能够较好地消除回声，因此残留信号中不会夹杂较多回声信号。因此参考信号和残留信号的相关性较小，因此对自适应滤波器的调整也较小。而实际情况中，由于自适应滤波器已进入稳定阶段，因此能够进行较好的回声消除，此时确实不需要对自适应滤波器系数进行较大调整，只需要稳定滤波器系数即可，本申请实施例对自适应滤波器系数的更新处理能满足实际需求。
第三种、近端单讲：即本端用户发出声音，远端用户没有发出声音的情况，对于本端来说，远端用户所传输的参考信号几乎为零，本端用户的麦克风主要采集本端用户的声音，而不会采集到较多回声信号，因此残留信号强度较强。此种情况下，参考信号几乎为零，而残留信号较强，因此两者相关系数也几乎为零，步长因子也几乎为零，无需对自适应滤波器系数进行过多调整，使自适应滤波器保持当前状态即可。而实际情况，由于远端用户没有发 出声音，麦克风不会采集到过多回声信号，确实无需进行回声消除。
第四种、双端讲话：即本端用户发出声音，远端用户也发出声音的情况。当从其他状态进入双端讲话状态，此时自适应滤波器需要及时更新系数。由于远端用户发出声音，参考信号较强；而本端用户也发出声音，因此残留信号强度也较强。由于参考信号和残留信号中都有两个用户发出的声音，此种情况下对自适应滤波器的调整效果可能没有前三种明显，但双端同时讲话的情况可能较少出现，本实施例方案仍可以满足大多数情况的自适应滤波器系数调整。
与前述回声消除方法的实施例相对应，本申请还提供了回声消除装置、会议平板和计算机存储介质的实施例。
本申请回声消除装置的实施例可以应用在电子设备上。装置实施例可以通过软件实现，也可以通过硬件或者软硬件结合的方式实现。以软件实现为例，作为一个逻辑意义上的装置，是通过其所在电子设备的处理器将非易失性存储器中对应的计算机程序指令读取到内存中运行形成的。从硬件层面而言，如图3所示，为本申请回声消除装置所在电子设备的一种硬件结构图，除了图3所示的处理器310、内存330、网络接口320、以及非易失性存储器340之外，实施例中装置331所在的电子设备通常根据该电子设备的实际功能，还可以包括其他硬件，对此不再赘述。
如图4所示，图4是本申请根据一示例性实施例示出的一种回声消除装置的框图，所述装置包括：
信号获取模块41，用于：获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号。
回声消除模块42，用于：利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出。
系数更新模块43，用于：根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
在一个可选的实现方式中，所述系数更新模块43，还用于：
计算用于指示所述残留信号与所述参考信号之间相关性的相关系数；
根据所述相关系数确定用于调整所述自适应滤波器系数的步长因子，利用所述步长因子更新所述自适应滤波器系数。
在一个可选的实现方式中，所述系数更新模块43，还用于：
计算所述残留信号在频域上的功率谱与所述参考信号在频域上的功率谱的相关系数。
在一个可选的实现方式中，所述系数更新模块43，还用于：
通过如下公式计算所述相关系数：
Figure PCTCN2017104391-appb-000007
其中，cohxe为所述相关系数，所述xPow(f)为所述参考信号在频域上的功率谱，所述ePow(f)为所述残留信号在频域上的功率谱，所述xePow(f)为所述xPow(f)与所述残留信号的共轭信号的相关功率谱。
在一个可选的实现方式中，所述回声消除模块42，包括：
确定所述参考信号在频域上的频域参考信号，所述频域参考信号中包括多个频点；
利用所述自适应滤波器系数计算所述频域参考信号中的每个频点对应的估计频点，获得所述回声信号。
在一个可选的实现方式中，所述相关系数包括：所述残留信号与所述参考信号在频域上的各个频点对应的相关系数；
所述系数更新模块43，还用于：
根据如下一种或多种方式确定用于调整所述自适应滤波器系数的步长因子：
将所述在频域上的各个频点对应的相关系数分别作为所述自适应滤波器系数在频域上各个频点的步长因子；
统计所述在频域上的各个频点对应的相关系数的平均值，将所述平均值作为所述自适应滤波器系数在频域上每个频点的步长因子；
确定所述在频域上的各个频点对应的相关系数的中位数，将所述中位数作为所述自适应滤波器系数在频域上每个频点的步长因子。
在一个可选的实现方式中，所述系数更新模块43，还用于：
通过如下公式更新所述自适应滤波器系数：
Figure PCTCN2017104391-appb-000008
其中，f为频点，Wk(f)为第k时刻针对频点f的自适应滤波器系数，μ(f)为频点f的步长因子，X(k)为所述频域参考信号，E(f)为所述残留信号在频域上的频域残留信号。
上述回声消除装置中各个模块的功能和作用的实现过程具体详见上述回声消除方法中对应步骤的实现过程，在此不再赘述。
对于装置实施例而言，由于其基本对应于方法实施例，所以相关之处参见方法实施例的部分说明即可。以上所描述的装置实施例仅仅是示意性的，其中所述作为分离部件说明的模块可以是或者也可以不是物理上分开的，作为模块显示的部件可以是或者也可以不是物理模块，即可以位于一个地方，或者也可以分布到多个网络模块上。可以根据实际的需要选择其中的部分或者全部模块来实现本申请方案的目的。本领域普通技术人员在不付出创造性劳动的情况下，即可以理解并实施。
相应的，本申请实施例还提供一种会议平板，所述会议平板包括回声消除装置，所述回声消除装置用于：
获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
相应的，本申请实施例一种计算机存储介质，所述存储介质中存储有程序指令，所述程序指令包括：
获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
本申请可采用在一个或多个其中包含有程序代码的存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。计算机可用存储介质包括永久性和非永久性、可移动和非可移动媒体，可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括但不限于：相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器 (EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带，磁带磁磁盘存储或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。
本领域技术人员在考虑说明书及实践这里申请的发明后，将容易想到本申请的其它实施方案。本申请旨在涵盖本申请的任何变型、用途或者适应性变化，这些变型、用途或者适应性变化遵循本申请的一般性原理并包括本申请未申请的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的，本申请的真正范围和精神由下面的权利要求指出。
应当理解的是，本申请并不局限于上面已经描述并在附图中示出的精确结构，并且可以在不脱离其范围进行各种修改和改变。本申请的范围仅由所附的权利要求来限制。
以上所述仅为本申请的较佳实施例而已，并不用以限制本申请，凡在本申请的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本申请保护的范围之内。

Claims (10)

  1. 一种回声消除方法，其特征在于，所述方法包括：
    获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
    利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
    根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数，包括：
    计算用于指示所述残留信号与所述参考信号之间相关性的相关系数；
    根据所述相关系数确定用于调整所述自适应滤波器系数的步长因子，利用所述步长因子更新所述自适应滤波器系数。
  3. 根据权利要求2所述的方法，其特征在于，所述计算用于指示所述残留信号与所述参考信号之间相关性的相关系数，包括：
    计算所述残留信号在频域上的功率谱与所述参考信号在频域上的功率谱的相关系数。
  4. 根据权利要求3所述的方法，其特征在于，所述计算所述残留信号在频域上的功率谱与所述参考信号在频域上的功率谱的相关系数，包括：
    通过如下公式计算所述相关系数：
    Figure PCTCN2017104391-appb-100001
    其中，cohxe为所述相关系数，所述xPow(f)为所述参考信号在频域上的功率谱，所述ePow(f)为所述残留信号在频域上的功率谱，所述xePow(f)为所述xPow(f)与所述残留信号的共轭信号的相关功率谱。
  5. 根据权利要求1所述的方法，其特征在于，所述利用所述自适应滤波器系数估计所述参考信号对应的回声信号，包括：
    确定所述参考信号在频域上的频域参考信号，所述频域参考信号中包括多个频点；
    利用所述自适应滤波器系数计算所述频域参考信号中的每个频点对应的估计频点，获得所述回声信号。
  6. 根据权利要求5所述的方法，其特征在于，所述相关系数包括：所述残留信号与所述参考信号在频域上的各个频点对应的相关系数；
    所述根据所述相关系数确定用于调整所述自适应滤波器系数的步长因子，包括如下一种或多种方式：
    将所述在频域上的各个频点对应的相关系数分别作为所述自适应滤波器系数在频域上各个频点的步长因子；
    统计所述在频域上的各个频点对应的相关系数的平均值，将所述平均值作为所述自适应滤波器系数在频域上每个频点的步长因子；
    确定所述在频域上的各个频点对应的相关系数的中位数，将所述中位数作为所述自适应滤波器系数在频域上每个频点的步长因子。
  7. 根据权利要求2所述的方法，其特征在于，所述根据所述步长因子更新所述自适应滤波器系数，包括：
    通过如下公式更新所述自适应滤波器系数：
    Figure PCTCN2017104391-appb-100002
    其中，f为频点，Wk(f)为第k时刻针对频点f的自适应滤波器系数，μ(f)为频点f的步长因子，X(k)为所述频域参考信号，E(f)为所述残留信号在频域上的频域残留信号。
  8. 一种回声消除装置，其特征在于，所述装置包括：
    信号获取模块，用于：获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
    回声消除模块，用于：利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
    系数更新模块，用于：根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
  9. 一种会议平板，其特征在于，所述会议平板包括回声消除装置，所述回声消除装置用于：
    获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
    利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
    根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
  10. 一种计算机存储介质，其特征在于，所述存储介质中存储有程序指令，所述程序指令包括：
    获取输入至扬声器进行播放的参考信号，以及获取麦克风的采集信号；
    利用自适应滤波器系数估计所述参考信号对应的回声信号，并从所述采集信号中消除所述回声信号，获得残留信号并输出；
    根据所述残留信号与所述参考信号之间的相关性更新所述自适应滤波器系数。
PCT/CN2017/104391 2017-04-12 2017-09-29 回声消除方法、装置、会议平板及计算机存储介质 WO2018188282A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710237357.5 2017-04-12
CN201710237357.5A CN107123430B (zh) 2017-04-12 2017-04-12 回声消除方法、装置、会议平板及计算机存储介质

Publications (1)

Publication Number Publication Date
WO2018188282A1 true WO2018188282A1 (zh) 2018-10-18

Family

ID=59725577

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/104391 WO2018188282A1 (zh) 2017-04-12 2017-09-29 回声消除方法、装置、会议平板及计算机存储介质

Country Status (2)

Country Link
CN (1) CN107123430B (zh)
WO (1) WO2018188282A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11869528B2 (en) 2019-03-14 2024-01-09 Suzhou Keda Technology Co., Ltd. Voice signal processing method and device, apparatus, and readable storage medium

Families Citing this family (37)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107123430B (zh) * 2017-04-12 2019-06-04 广州视源电子科技股份有限公司 回声消除方法、装置、会议平板及计算机存储介质
CN107134281A (zh) * 2017-05-04 2017-09-05 重庆第二师范学院 一种自适应回声消除中自适应滤波器系数更新方法
CN109524018B (zh) * 2017-09-19 2022-06-10 华为技术有限公司 一种回声处理方法及设备
CN107958671B (zh) * 2017-11-22 2021-04-13 广州市六体智能科技有限公司 一种执法记录仪的回声消除装置
CN109961798B (zh) * 2017-12-26 2021-06-11 华平信息技术股份有限公司 回声消除系统、方法、可读计算机存储介质、及终端
CN108550371B (zh) * 2018-03-30 2021-06-01 云知声智能科技股份有限公司 智能语音交互设备快速稳定的回声消除方法
CN108289267A (zh) * 2018-04-14 2018-07-17 北京智网时代科技有限公司 消除电视干扰的回声消除装置、方法、音箱、音频发送器
CN108630219B (zh) * 2018-05-08 2021-05-11 北京小鱼在家科技有限公司 回声抑制音频信号特征跟踪的处理系统、方法及装置
CN110164446B (zh) * 2018-06-28 2023-06-30 腾讯科技（深圳）有限公司 语音信号识别方法和装置、计算机设备和电子设备
CN109087665B (zh) * 2018-07-06 2022-11-11 南京时保联信息科技有限公司 一种非线性回声抑制方法
CN109102821B (zh) * 2018-09-10 2021-05-25 思必驰科技股份有限公司 时延估计方法、系统、存储介质及电子设备
WO2020063798A1 (zh) * 2018-09-27 2020-04-02 深圳市冠旭电子股份有限公司 一种回声消除方法、装置及智能音箱
CN109559756B (zh) * 2018-10-26 2021-05-14 北京佳讯飞鸿电气股份有限公司 滤波系数确定方法、回声消除方法、相应装置及设备
CN110021289B (zh) * 2019-03-28 2021-08-31 腾讯科技（深圳）有限公司 一种声音信号处理方法、装置及存储介质
CN110211599B (zh) * 2019-06-03 2021-07-16 Oppo广东移动通信有限公司 应用唤醒方法、装置、存储介质及电子设备
CN110148421B (zh) * 2019-06-10 2021-07-20 浙江大华技术股份有限公司 一种残余回声检测方法、终端和装置
CN110265054B (zh) * 2019-06-14 2024-01-30 深圳市腾讯网域计算机网络有限公司 语音信号处理方法、装置、计算机可读存储介质和计算机设备
WO2020252629A1 (zh) * 2019-06-17 2020-12-24 深圳市汇顶科技股份有限公司 残余回声检测方法、残余回声检测装置、语音处理芯片及电子设备
US11017792B2 (en) * 2019-06-17 2021-05-25 Bose Corporation Modular echo cancellation unit
CN110600048B (zh) * 2019-08-23 2022-03-25 Oppo广东移动通信有限公司 音频校验方法、装置、存储介质及电子设备
CN112447175A (zh) * 2019-08-29 2021-03-05 北京声智科技有限公司 一种回波消除方法和装置
CN110853664B (zh) * 2019-11-22 2022-05-06 北京小米移动软件有限公司 评估语音增强算法性能的方法及装置、电子设备
CN112929506B (zh) * 2019-12-06 2023-10-17 阿里巴巴集团控股有限公司 音频信号的处理方法及装置，计算机存储介质及电子设备
CN110992923B (zh) * 2019-12-19 2022-06-07 浙江大华技术股份有限公司 回声消除方法、电子设备以及存储装置
CN111048061B (zh) * 2019-12-27 2022-12-27 西安讯飞超脑信息科技有限公司 回声消除滤波器的步长获取方法、装置及设备
CN111370015B (zh) * 2020-02-28 2021-03-19 北京字节跳动网络技术有限公司 回声消除方法、装置、电子设备及存储介质
CN111199748B (zh) * 2020-03-12 2022-12-27 紫光展锐(重庆)科技有限公司 回声消除方法、装置、设备以及存储介质
CN111445917B (zh) * 2020-03-17 2023-02-10 浙江大华技术股份有限公司 一种回声消除方法、装置以及计算机存储介质
CN111724804A (zh) * 2020-06-29 2020-09-29 北京百度网讯科技有限公司 用于处理信息的方法和装置
CN111885276B (zh) * 2020-07-31 2021-05-25 支付宝(杭州)信息技术有限公司 消除回声的方法及系统
CN112017679B (zh) * 2020-08-05 2024-01-26 海尔优家智能科技（北京）有限公司 用于自适应滤波器系数更新的方法及装置、设备
CN111916098A (zh) * 2020-08-13 2020-11-10 科大讯飞(苏州)科技有限公司 语音信号处理方法、装置及设备
TWI759858B (zh) * 2020-09-11 2022-04-01 點通科技股份有限公司 具回音消除之通話系統及其操作方法
CN112397080B (zh) * 2020-10-30 2023-02-28 浙江大华技术股份有限公司 回声消除方法及装置、语音设备及计算机可读存储介质
CN113055787B (zh) * 2021-03-25 2022-10-04 科大讯飞股份有限公司 回声消除方法、装置、电子设备和存储介质
CN113808609A (zh) * 2021-09-18 2021-12-17 展讯通信（上海）有限公司 回声检测方法及装置、计算机可读存储介质、终端设备
CN116110393B (zh) * 2023-02-01 2024-01-23 镁佳(北京)科技有限公司 一种基于语音相似度的拒识方法、装置、计算机及介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001077730A (ja) * 1999-09-07 2001-03-23 Fujitsu Ltd 適応フィルタの係数推定装置
US6792106B1 (en) * 1999-09-17 2004-09-14 Agere Systems Inc. Echo canceller and method of echo cancellation using an NLMS algorithm
CN1953060A (zh) * 2006-11-24 2007-04-25 北京中星微电子有限公司 一种麦克风回声消除装置及回声消除方法
CN104754157A (zh) * 2013-12-26 2015-07-01 联芯科技有限公司 一种残留回声抑制方法和系统
CN107123430A (zh) * 2017-04-12 2017-09-01 广州视源电子科技股份有限公司 回声消除方法、装置、会议平板及计算机存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001077730A (ja) * 1999-09-07 2001-03-23 Fujitsu Ltd 適応フィルタの係数推定装置
US6792106B1 (en) * 1999-09-17 2004-09-14 Agere Systems Inc. Echo canceller and method of echo cancellation using an NLMS algorithm
CN1953060A (zh) * 2006-11-24 2007-04-25 北京中星微电子有限公司 一种麦克风回声消除装置及回声消除方法
CN104754157A (zh) * 2013-12-26 2015-07-01 联芯科技有限公司 一种残留回声抑制方法和系统
CN107123430A (zh) * 2017-04-12 2017-09-01 广州视源电子科技股份有限公司 回声消除方法、装置、会议平板及计算机存储介质

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11869528B2 (en) 2019-03-14 2024-01-09 Suzhou Keda Technology Co., Ltd. Voice signal processing method and device, apparatus, and readable storage medium

Also Published As

Publication number Publication date
CN107123430A (zh) 2017-09-01
CN107123430B (zh) 2019-06-04

Similar Documents

Publication Publication Date Title
WO2018188282A1 (zh) 回声消除方法、装置、会议平板及计算机存储介质
US11297178B2 (en) Method, apparatus, and computer-readable media utilizing residual echo estimate information to derive secondary echo reduction parameters
JP4955228B2 (ja) ラウンドロビン正則化を用いたマルチチャネルエコーキャンセレーション
Enzner et al. Acoustic echo control
US10978086B2 (en) Echo cancellation using a subset of multiple microphones as reference channels
US9286883B1 (en) Acoustic echo cancellation and automatic speech recognition with random noise
CN110176244B (zh) 回声消除方法、装置、存储介质和计算机设备
CN110782914B (zh) 信号处理方法、装置、终端设备及存储介质
JP2016518628A (ja) マルチチャネルエコー消去および雑音抑圧
US20160006880A1 (en) Variable step size echo cancellation with accounting for instantaneous interference
CN105432062B (zh) 用于回波去除的方法、设备及介质
CN110431624B (zh) 残余回声检测方法、残余回声检测装置、语音处理芯片及电子设备
CN112929506B (zh) 音频信号的处理方法及装置，计算机存储介质及电子设备
Tashev Recent advances in human-machine interfaces for gaming and entertainment
Cohen et al. An online algorithm for echo cancellation, dereverberation and noise reduction based on a Kalman-EM Method
Gueraini et al. New variable step-size fast NLMS algorithm for non-stationary systems
US10540984B1 (en) System and method for echo control using adaptive polynomial filters in a sub-band domain
Fukui et al. Acoustic echo and noise canceller for personal hands-free video IP phone
CN111989934B (zh) 回声消除装置、回声消除方法、信号处理芯片及电子设备
CN112489680A (zh) 声学回声消除算法的评估方法、装置及终端设备
Garre et al. An Acoustic Echo Cancellation System based on Adaptive Algorithm
Ruiz et al. Distributed combined acoustic echo cancellation and noise reduction using GEVD-based distributed adaptive node specific signal estimation with prior knowledge
JP4425114B2 (ja) 反響消去方法、反響消去装置、反響消去プログラム、及びこれを記録した記録媒体
US20230410828A1 (en) Systems and methods for echo mitigation
CN115631761A (zh) 回声消除装置、方法、计算机设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17905868

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 17.02.2020)

122 Ep: pct application non-entry in european phase

Ref document number: 17905868

Country of ref document: EP

Kind code of ref document: A1