CN107564539B - Acoustic echo cancellation method and device facing microphone array - Google Patents
Acoustic echo cancellation method and device facing microphone array Download PDFInfo
- Publication number
- CN107564539B CN107564539B CN201710757604.4A CN201710757604A CN107564539B CN 107564539 B CN107564539 B CN 107564539B CN 201710757604 A CN201710757604 A CN 201710757604A CN 107564539 B CN107564539 B CN 107564539B
- Authority
- CN
- China
- Prior art keywords
- signal
- echo cancellation
- module
- microphone
- microphone array
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
Images
Abstract
The invention relates to a microphone array-oriented acoustic echo cancellation method, which comprises the steps that a microphone array collects data of near-end signals and echo signals in real time, channel signals of one microphone are selected, an initialization model is built according to a self-adaptive filtering thought and an independent component analysis algorithm, an echo cancellation model is built according to the initialization model to perform echo cancellation on single-channel signals of the microphone, and echo cancellation estimation models of other microphones in the microphone array are built according to the initial echo cancellation model and the filter coefficients of single channels to perform echo cancellation; the device provided by the invention performs echo cancellation on the multi-channel recording signal and the echo signal of the measuring system by arranging the acoustic sensor at the key position on the annular sensor array, so as to obtain the echo cancellation signal of the microphone array, and finally achieve the purpose of improving the voice recognition rate.
Description
Technical Field
The invention relates to the field of digital signal processing, in particular to an acoustic echo cancellation method and device for a microphone array.
Background
With the advent of the artificial intelligence era, the voice technology has received more and more attention as an interface for human-computer interaction. The traditional near-field voice interaction technology cannot meet the requirements of people, and people hope to develop intelligent equipment capable of controlling voice in more distant and complex scenes. Thus, microphone array technology becomes the core of far-field speech interaction.
By microphone array is meant a system of a plurality of microphones in which a certain number of acoustic sensors (microphones) are arranged in a certain regular pattern and applied to the processing of speech signals. In the traditional voice interaction application scene, a single microphone is mainly adopted, but a single microphone system can obtain a signal meeting the voice recognition requirement only under the near conditions of close sound source, low noise and no reverberation, so that the single microphone system has certain limitation. For the application scenario of far-field speech recognition, the quality of signals picked up by a microphone system is poor due to a large amount of noise, reverberation and echo contained in the real environment, and the recognition rate of speech is seriously affected. And due to the problems of multiple sound sources and environmental noise, the single-microphone system is difficult to separate and locate the sound sources, so the multi-microphone system is required to assist the front-end signal processing of speech recognition.
Aiming at the current complex application scene, a series of key technologies capable of effectively improving the speech recognition rate are developed based on a microphone array, and the key technologies mainly comprise: speech enhancement, sound source localization, reverberation cancellation, echo cancellation, noise suppression. The voice enhancement technology mainly adopts a beam forming method, which mainly enhances sound in a specific direction and inhibits sound interference except a main lobe; sound source positioning, which mainly refers to calculating the angle and distance of a target speaker by using a microphone array so as to realize the tracking of the target speaker and the directional pickup of sound; reverberation elimination, which mainly utilizes signal preprocessing means such as inverse filtering or linear prediction to eliminate the phenomenon of mutual superposition of asynchronous voices influencing voice recognition; echo cancellation, which mainly utilizes means such as adaptive signal processing to eliminate phenomena such as accent in interference speech recognition; and noise suppression, wherein noise reduction methods such as spectral subtraction are mainly used for eliminating the situation of large background noise in a speech recognition application scene.
Based on the engineering application background, echo cancellation is taken as one of key technologies of a microphone array, and important technical support can be provided for far-field voice interaction for a multi-channel acoustic echo cancellation algorithm under the microphone array.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a microphone array-oriented acoustic echo cancellation method and device based on the combination of an independent component analysis algorithm and a self-adaptive filtering idea, aiming at the problem that acoustic echo cancellation is difficult in far-field speech recognition.
According to the acoustic echo cancellation method facing the microphone array, the microphone array collects data of near-end signals and echo signals in real time;
selecting a microphone channel signal output by one of the microphones, constructing an initialization model according to a self-adaptive filtering thought and an independent component analysis algorithm, and establishing an echo cancellation model according to the initialization model to perform echo cancellation on the microphone channel signal;
and constructing echo cancellation estimation models of other microphones in the microphone array according to the initial echo cancellation model and the single-channel filter coefficient for echo cancellation.
Further, the microphone channel signal includes a near-end signal and an echo signal collected by the microphone.
Further, the initialization model includes a functional relationship between the echo signal and the far-end signal in the microphone channel signal and a functional relationship between the echo signal and the far-end signal estimated in the other microphone channel signals as
wherein
An impulse response of a far-end signal propagating through space,
in order to be an echo signal, the echo signal,
in order to estimate the echo signal, it is,
in order to be a far-end signal,
is the estimated adaptive filter coefficients.
Further, the relationship between the estimated echo signal and the error signal for controlling the convergence of the adaptive filter is
Further, the echo cancellation model is
Further, the echo cancellation estimation model is
More particularly, the
The solution was performed by Fast ICA.
The invention provides an acoustic echo cancellation device facing a microphone array, which comprises a central processing module, an acquisition module, a storage module, a power supply module, an upper computer communication module, a reset module and a key control module, wherein the central processing module is used for controlling the acquisition module and processing signals acquired by the acquisition module, the storage module is used for storing signals input by a central processing unit into the storage module and signals output by the storage module to the central processing module, the upper computer communication module is used for processing the signals processed by the central processing module and controlling the central processing module, the reset module is used for controlling the reset of the central processing module, and the key control module is used for inputting parameters into the central processing module. The acoustic echo cancellation method and device for the microphone array can achieve echo cancellation under multiple channels, and improve voice recognition rate.
Furthermore, the central processing unit module is an OMAP dual-core processor based on an ARM core and a DSP core.
Further, the acquisition module is a microphone array data acquisition module.
Further, the upper computer communication module further comprises a cloud server for voice recognition.
The acoustic echo cancellation method and device for the microphone array can effectively perform multi-channel echo cancellation on a multi-microphone array system, provide effective preprocessing algorithm technical support for rear-end speech recognition, and are applied to devices or systems such as intelligent robots, intelligent sound boxes and intelligent conferences.
The foregoing description is only an overview of the technical solutions of the present invention, and in order to make the technical solutions of the present invention more clearly understood and to implement them in accordance with the contents of the description, the following detailed description is given with reference to the preferred embodiments of the present invention and the accompanying drawings.
Drawings
Fig. 1 is a functional block diagram of an algorithm of an acoustic echo cancellation method for a microphone array according to the present invention;
fig. 2 is a block diagram of an acoustic echo cancellation device facing a microphone array according to the present invention.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
Referring to fig. 1, fig. 1 shows a preferred embodiment of an acoustic echo cancellation method facing a microphone array according to the present invention. In this embodiment:
firstly, a microphone array collects data of a near-end signal in a recording channel and an echo signal in an echo channel in real time, then a microphone channel signal output by one of the microphones is selected, an initialization model of an ICA (independent component analysis) is built according to the data in the microphone channel signal and an adaptive filtering thought and an independent component analysis algorithm, an echo cancellation model is built according to the initialization model of the ICA to perform echo cancellation on the microphone channel signal, and an echo cancellation estimation model of other microphones in the microphone array is built according to the initial echo cancellation model and a filter coefficient of a single channel to perform echo cancellation.
Preferably, the method comprises the following specific steps:
s1, setting the measured far-end signal as
Signal D collected by microphone arraykIs composed of
Wherein m represents the number of acoustic sensors arranged on the microphone structure, k represents the number of time domain sampling points, and j is more than or equal to 1 and less than or equal to m; i is more than or equal to 1 and less than or equal to k;
s2, setting the near-end voice signal as
The echo signal is
Therefore, the j-th microphone channel signal, the near-end signal and the echo signal satisfy the following relations:
s3, setting the impulse response of the far-end signal through space propagation as
Echo signal
And remote end signal
The relationship of (1) is:
s4, setting the estimated echo signal as
The estimated adaptive filter coefficient is
Error signal for estimating relation between echo signal and far-end signal and controlling adaptive filtering convergence
The relationship is as follows:
s5, constructing an acoustic echo cancellation ICA model based on the ICA idea, wherein the model equation can be expressed as:
wherein the content of the first and second substances,
in order to obtain the desired original signal,
in order to observe the vector, the vector is,
is a mixing matrix;
s6, an equation for separating the original signals can be established according to S05, which can be expressed as follows:
wherein the content of the first and second substances,
is a non-0 scalar to obtain
S7, in order to separate the target voice as much as possible, Fast ICA method is adopted to solve
The method comprises the following specific steps:
(a) to observation data matrix
Carrying out equalization;
(b) is provided with
Whitening observation data matrix x (Z ═ vx), where v ═ D-1/2ETD, E are feature values and feature vectors corresponding to the covariance matrix of the observation data matrix x respectively;
(c) selecting the number of components to be separated of ICA, and setting a counter l to be 1;
(d) setting initial random weight coefficients
(e) Update solving
(f) Decorrelation:
(g) normalization weight coefficient:
(h) if it is not
Non-convergence
Returning to the step (e);
(i) iteration l ═ l +1, if l ≦ m, return to step (d);
s8, for the j +1 th channel, the weight coefficient obtained by S07
And approximating echo cancellation of other channels, wherein the equation is as follows:
the estimated calculation formula for the near-end signals in the other microphone channel signals is as follows:
thus, the method utilizes Independent Component Analysis (ICA) algorithm and adaptive filtering idea to construct an echo cancellation estimation model, separates target voice, namely near-end signal, from far-end signal and microphone pick-up signal, and finally transmits the separated target voice to a recognition end, thereby achieving the purposes of echo cancellation and voice recognition.
Referring to fig. 2, fig. 2 is a preferred embodiment of an acoustic echo cancellation device facing a microphone array according to the present invention, for implementing the above-mentioned acoustic echo cancellation method facing the microphone array. In this embodiment:
the acoustic echo cancellation device facing the microphone array comprises a central processing module, an acquisition module, a storage module, a power supply module, an upper computer communication module, a reset module and a keying module, wherein the central processing module is used for controlling the acquisition module and processing signals acquired by the acquisition module, the storage module is used for storing signals input into the storage module by a central processing unit and signals output to the central processing module by the storage module, the upper computer communication module is used for processing signals processed by the central processing module and controlling the central processing module, the reset module is used for controlling the reset of the central processing module, and the keying module is used for inputting parameters into the central processing module.
Preferably, in order to make the central processing module have low power consumption and strong data processing capability, the central processing module is an OMAP dual-core processor having a dual-core ARM core and a DSP core structure.
Preferably, the acquisition module is a microphone array data acquisition module having functions of signal input, signal conditioning, and a/D data acquisition and conversion.
As preferred, host computer communication module still includes the high in the clouds server that is used for carrying out speech recognition, like this, host computer communication module can carry out speech recognition with the data after the echo cancellation's discernment end on the high in the clouds server of network transmission.
The working steps of the acoustic echo cancellation device facing the microphone array are as follows:
firstly, parameter setting is carried out on the working state, channel control, signal type acquisition, sampling frequency and the like of the device through an upper computer communication module or a keying module, a parameter instruction is transmitted to a central processing module, and a driving acquisition module acquires voice data; then, the DSP module of the OMAP processor performs echo cancellation on the echo channel signals acquired by the microphone array and the mixed signals acquired by the recording channel; and finally, transmitting the data after echo cancellation to a voice recognition system of a cloud server of the upper computer communication module through a network.
The acoustic echo cancellation device facing the microphone array arranges a plurality of acoustic sensors on key points of a ring-shaped microphone array structure, performs echo cancellation on a voice signal obtained by measurement in real time, performs echo cancellation on a system, and applies the acoustic echo cancellation device to an intelligent sound box or an intelligent robot. The multichannel echo cancellation device takes an OMAP embedded processor as a core and integrates units for data acquisition, data storage, control, data processing and the like, the device is designed to fully utilize dual-core ARM core and DSP core structures of the OMAP, and has the functions of low ARM core power consumption, high processing speed, flexible task scheduling and the like and the function of powerful digital processing analysis of the DSP core, and the two are effectively combined to realize real-time online acquisition, processing, transmission and analysis of voice signals. Meanwhile, the Ethernet is adopted for data transmission, so that the data is transmitted quickly and efficiently, the loss of signals in transmission is avoided, resource sharing is realized, and the defects of offline and delay of traditional data acquisition are overcome.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, it should be noted that, for those skilled in the art, many modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (4)
1. An acoustic echo cancellation method facing a microphone array is characterized in that:
the microphone array collects near-end signals and echo signal data in real time;
selecting a microphone channel signal acquired by one microphone, constructing an initialization model according to a self-adaptive filtering thought and an independent component analysis algorithm, and establishing an echo cancellation model according to the initialization model to perform echo cancellation on the microphone channel signal; according to the initial echo cancellation model and the filter coefficient of the single channel, constructing echo cancellation estimation models of other microphones in the microphone array for echo cancellation;
the microphone channel comprises a near-end signal and an echo signal collected by the microphone;
the method specifically comprises the following steps: s1, setting the measured far-end signal as
Signal D collected by microphone arraykComprises the following steps:
wherein m represents the number of acoustic sensors arranged on the microphone structure, k represents the number of time domain sampling points, and j is more than or equal to 1 and less than or equal to m; i is more than or equal to 1 and less than or equal to k;
s2, setting the near-end voice signal as
The echo signal is
Therefore, the j-th microphone channel signal, the near-end signal and the echo signal satisfy the following relations:
s3, setting the impulse response of the far-end signal through space propagation as
Echo signal
And remote end signal
The relationship of (1) is:
s4, setting the estimated echo signal as
The estimated adaptive filter coefficient is
Error signal for estimating relation between echo signal and far-end signal and controlling adaptive filtering convergence
The relationship is as follows:
s5, constructing an acoustic echo cancellation ICA model based on the ICA idea, wherein the model equation can be expressed as:
wherein the content of the first and second substances,
in order to obtain the desired original signal,
in order to observe the vector, the vector is,
is a mixing matrix;
s6, an equation for separating the original signals can be established according to S5, which can be expressed as follows:
wherein the content of the first and second substances,
is a non-0 scalar to obtain
S7 solving by Fast ICA method
In order to separate the target voice, the specific steps are as follows:
(a) to observation data matrix
Carrying out equalization;
(b) is provided with
Whitening observation data matrix x (Z ═ vx), where v ═ D-1/2ETD, E are feature values and feature vectors corresponding to the covariance matrix of the observation data matrix x respectively;
(c) selecting the number of components to be separated of ICA, and setting a counter l to be 1;
(d) setting initial random weight coefficients
(e) Update solving
(f) Decorrelation:
(g) normalization weight coefficient:
(h) if it is not
Non-convergence
Returning to the step (e);
(i) iteration l ═ l +1, if l ≦ m, return to step (d);
s8, for the j +1 th channel, the weight coefficient obtained by S07
And approximating echo cancellation of other channels, wherein the equation is as follows:
the estimated calculation formula for the near-end signals in the other microphone channel signals is as follows:
2. an acoustic echo cancellation device facing a microphone array, characterized by: including central processing module, collection module, storage module, power module, host computer communication module, module and keying module reset, central processing module is used for control collection module and processing the signal that collection module gathered, storage module is used for the storage central processing module input storage module's signal and storage module to the signal that central processing module exported, host computer communication module is used for handling the warp the signal that central processing module handled and control central processing module, the module that resets is used for controlling central processing module's reset, keying module be used for to central processing module input parameter adopts claim 1 the acoustic echo cancellation method towards microphone array eliminate the echo.
3. The microphone array-facing acoustic echo cancellation device of claim 2, wherein: the central processing module is an OMAP dual-core processor based on an ARM core and a DSP core, and the acquisition module is a microphone array data acquisition module.
4. An acoustic echo cancellation device facing a microphone array according to claim 3, characterized in that: the upper computer communication module further comprises a cloud server for voice recognition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710757604.4A CN107564539B (en) | 2017-08-29 | 2017-08-29 | Acoustic echo cancellation method and device facing microphone array |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710757604.4A CN107564539B (en) | 2017-08-29 | 2017-08-29 | Acoustic echo cancellation method and device facing microphone array |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107564539A CN107564539A (en) | 2018-01-09 |
| CN107564539B true CN107564539B (en) | 2021-12-28 |
Family
ID=60977477
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710757604.4A Active CN107564539B (en) | 2017-08-29 | 2017-08-29 | Acoustic echo cancellation method and device facing microphone array |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107564539B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108305631B (en) * | 2018-04-04 | 2023-09-05 | 西安合谱声学科技有限公司 | Acoustic processing equipment based on multi-core modularized framework |
| CN110782911A (en) * | 2018-07-30 | 2020-02-11 | 阿里巴巴集团控股有限公司 | Audio signal processing method, apparatus, device and storage medium |
| CN111261179A (en) * | 2018-11-30 | 2020-06-09 | 阿里巴巴集团控股有限公司 | Echo cancellation method and device and intelligent equipment |
| CN110265054B (en) * | 2019-06-14 | 2024-01-30 | 深圳市腾讯网域计算机网络有限公司 | Speech signal processing method, device, computer readable storage medium and computer equipment |
| CN111031448B (en) * | 2019-11-12 | 2021-09-17 | 西安讯飞超脑信息科技有限公司 | Echo cancellation method, echo cancellation device, electronic equipment and storage medium |
| CN111312269B (en) * | 2019-12-13 | 2023-01-24 | 天津职业技术师范大学(中国职业培训指导教师进修中心) | Rapid echo cancellation method in intelligent loudspeaker box |
| CN115696140B (en) * | 2022-12-05 | 2023-05-26 | 长沙东玛克信息科技有限公司 | Classroom audio multichannel echo cancellation method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1750573A (en) * | 2004-09-03 | 2006-03-22 | 哈曼贝克自动系统股份有限公司 | Speech signal processing with combined noise reduction and echo compensation |
| CN101820302A (en) * | 2009-02-27 | 2010-09-01 | 比亚迪股份有限公司 | Device and method for canceling echo |
| CN103426436A (en) * | 2012-05-04 | 2013-12-04 | 索尼电脑娱乐公司 | Source separation by independent component analysis in conjuction with optimization of acoustic echo cancellation |
| CN104883462A (en) * | 2014-02-28 | 2015-09-02 | 中国科学院声学研究所 | Adaptive filter and filtering method used for eliminating acoustics echo |
| CN106875938A (en) * | 2017-03-10 | 2017-06-20 | 南京信息工程大学 | A kind of improved nonlinear adaptive sound end detecting method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5375400B2 (en) * | 2009-07-22 | 2013-12-25 | ソニー株式会社 | Audio processing apparatus, audio processing method and program |
-
2017
- 2017-08-29 CN CN201710757604.4A patent/CN107564539B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1750573A (en) * | 2004-09-03 | 2006-03-22 | 哈曼贝克自动系统股份有限公司 | Speech signal processing with combined noise reduction and echo compensation |
| CN101820302A (en) * | 2009-02-27 | 2010-09-01 | 比亚迪股份有限公司 | Device and method for canceling echo |
| CN103426436A (en) * | 2012-05-04 | 2013-12-04 | 索尼电脑娱乐公司 | Source separation by independent component analysis in conjuction with optimization of acoustic echo cancellation |
| CN104883462A (en) * | 2014-02-28 | 2015-09-02 | 中国科学院声学研究所 | Adaptive filter and filtering method used for eliminating acoustics echo |
| CN106875938A (en) * | 2017-03-10 | 2017-06-20 | 南京信息工程大学 | A kind of improved nonlinear adaptive sound end detecting method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107564539A (en) | 2018-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107564539B (en) | Acoustic echo cancellation method and device facing microphone array | |
| CN107993670B (en) | Microphone array speech enhancement method based on statistical model | |
| CN106782584B (en) | Audio signal processing device, method and electronic device | |
| Zhang et al. | Why does PHAT work well in lownoise, reverberative environments? | |
| JP4654777B2 (en) | Acoustic echo cancellation device | |
| CN111161751A (en) | Distributed microphone pickup system and method under complex scene | |
| Silverman et al. | The huge microphone array | |
| US20180350379A1 (en) | Multi-Channel Speech Signal Enhancement for Robust Voice Trigger Detection and Automatic Speech Recognition | |
| US6826284B1 (en) | Method and apparatus for passive acoustic source localization for video camera steering applications | |
| US20170251301A1 (en) | Selective audio source enhancement | |
| US20220223149A1 (en) | Narrowband direction of arrival for full band beamformer | |
| CN108109617A (en) | A kind of remote pickup method | |
| RU2685053C2 (en) | Estimating room impulse response for acoustic echo cancelling | |
| CN102204276A (en) | Sound emission and collection device, and sound emission and collection method | |
| WO2022012206A1 (en) | Audio signal processing method, device, equipment, and storage medium | |
| Papp et al. | Hands-free voice communication with TV | |
| CN106782592B (en) | System and method for eliminating echo and howling of network sound transmission | |
| US20230197096A1 (en) | Audio signal processing method, training method, apparatus and storage medium | |
| CN110444220B (en) | Multi-mode remote voice perception method and device | |
| Hosseini et al. | Time difference of arrival estimation of sound source using cross correlation and modified maximum likelihood weighting function | |
| Tourbabin et al. | Speaker localization by humanoid robots in reverberant environments | |
| Hassani et al. | Distributed node-specific direction-of-arrival estimation in wireless acoustic sensor networks | |
| CN108650593B (en) | Three-microphone array far-field pickup method for teleconference | |
| Choi et al. | Convolutional neural network-based direction-of-arrival estimation using stereo microphones for drone | |
| CN113223552B (en) | Speech enhancement method, device, apparatus, storage medium, and program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |