CN107483761B - A kind of echo suppressing method and device - Google Patents
A kind of echo suppressing method and device Download PDFInfo
- Publication number
- CN107483761B CN107483761B CN201610399409.4A CN201610399409A CN107483761B CN 107483761 B CN107483761 B CN 107483761B CN 201610399409 A CN201610399409 A CN 201610399409A CN 107483761 B CN107483761 B CN 107483761B
- Authority
- CN
- China
- Prior art keywords
- echo
- linear
- signal
- audio signal
- component
- 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
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
- H04M9/082—Two-way loud-speaking telephone systems with means for conditioning the signal, e.g. for suppressing echoes for one or both directions of traffic using echo cancellers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R3/00—Circuits for transducers, loudspeakers or microphones
- H04R3/02—Circuits for transducers, loudspeakers or microphones for preventing acoustic reaction, i.e. acoustic oscillatory feedback
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Otolaryngology (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
- Telephone Function (AREA)
Abstract
The invention discloses a kind of echo suppressing method and devices.The method of the present invention includes: to inhibit to the linear echo component in the first audio signal, obtains the first echo suppressed signal;According to the first echo suppressed signal, the non linear echo components in the second audio signal are inhibited, obtain the second echo suppressed signal;Wherein, the first audio signal and the second audio signal are that the audio receiver in audio receiver array collects;Linear echo component in second echo suppressed signal is inhibited respectively with residual non linear echo components.The present invention can effectively inhibit linear and nonlinear acoustic echo.
Description
Technical field
The present invention relates to field of communication technology more particularly to a kind of echo suppressing method and devices.
Background technique
For mobile phone and tablet computer etc. have the equipment of communication function, hand-free call is usually that required application is special
One of property.For handsfree talk mode, acoustic echo suppression technology be for improving speech quality important technology, and by
Biggish technological challenge is brought to acoustic echo suppression technology again in the characteristic of acoustic echo real-time change.The prior art
In, there is the function that real-time change signal is quickly tracked in circumstances not known in view of sef-adapting filter, therefore usually in echo
It is used for valuation echo in suppression technology.
Due to the loudspeaker scale on mobile phone and tablet computer is smaller and with microphone at a distance of relatively close, and issued from loudspeaker
Sound it is usual and larger under handsfree talk mode so that at microphone pick up echo-signal compare near-end voice signals often
It is also strong;In addition, small scale loudspeaker when issuing loud sound and echo path upper-part vibration all caused it is non-thread
Property acoustic echo.This makes the existing monophonic acoustics echo killer realized based on linear adaptive filter, can not
Effectively eliminate and inhibit acoustic echo of such communication equipment under handsfree talk mode.
For acoustic echo nonlinear characteristic of the communication equipments such as mobile phone under handsfree talk mode, existing one
Nonlinear echoes suppressor is correspondingly used in a little technologies, but these technical solutions are larger in nonlinear residual echo component
When, it will inevitably inhibit when inhibiting nonlinear residual echo and then damage near-end voice signals, especially non-
When the amplitude of linear echo is suitable with near-end voice signals amplitude, using proximal end language in the technical solution of nonlinear echoes suppressor
Speech signal will will receive larger damage, at intended recipient end, received by voice signal be possibly even distorted completely.
Therefore, a kind of echo suppression techniques that can effectively inhibit linear and nonlinear acoustic echo how to be realized, are
The problem of industry is urgently studied and is explored.
Summary of the invention
The embodiment of the present invention provides a kind of echo suppressing method and device, to realize that one kind can effectively inhibit linearly
With the echo suppression techniques of nonlinear acoustics echo.
The echo suppressing method that some embodiments of the present invention provide, comprising:
Linear echo component in first audio signal is inhibited, the first echo suppressed signal is obtained;
According to first echo suppressed signal, the non linear echo components in the second audio signal are inhibited, are obtained
To the second echo suppressed signal;Wherein, first audio signal and second audio signal are in audio receiver array
Audio receiver collect;
Linear echo component in second echo suppressed signal is inhibited with residual non linear echo components.
In some alternative embodiments of the invention, the linear echo component in the first audio signal is inhibited, is obtained
First echo suppressed signal, comprising:
Far-end audio signal is filtered using the first sef-adapting filter to obtain the first linear echo component;
According to first linear echo component, the linear echo component in first audio signal is inhibited,
Obtain the first echo suppressed signal.
In some alternative embodiments of the invention, further includes: if judgement, which is currently at, is singly said state, certainly to described first
The coefficient vector of adaptive filter is updated.
In some alternative embodiments of the invention, according to first echo suppressed signal, in the second audio signal
Non linear echo components are inhibited, and the second echo suppressed signal is obtained, comprising:
First echo suppressed signal is filtered to obtain the first nonlinear echoes using the second sef-adapting filter
Component;
According to first non linear echo components, the non linear echo components in second audio signal are pressed down
System, obtains the second echo suppressed signal.
In some alternative embodiments of the invention, further includes: if judgement, which is currently at, is singly said state and first echo
The mean power of signal is inhibited to be greater than the first predetermined threshold value, or judgement is currently at and singly says state and the first linear echo point
The mean power of amount is greater than the second predetermined threshold value, then is updated to the coefficient vector of second sef-adapting filter, institute
It states the first linear echo component and far-end audio signal is filtered to obtain using the first sef-adapting filter.
In some alternative embodiments of the invention, to the linear echo component and residual in second echo suppressed signal
Non linear echo components are inhibited, comprising:
Far-end audio signal is filtered using third sef-adapting filter to obtain the second linear echo component;
According to second linear echo component, the linear echo component in second echo suppressed signal is pressed down
System, obtains third echo suppressed signal;
According to second echo suppressed signal and the third echo suppressed signal, nonlinear echoes suppressor pair is used
Residual non linear echo components in the third echo suppressed signal are inhibited.
In some alternative embodiments of the invention, further includes: if judgement, which is currently at, is singly said state, certainly to the third
The coefficient vector of adaptive filter is updated.
In some alternative embodiments of the invention, current whether be in of judgement singly says state in the following ways:
According to first audio signal and the first linear echo component, first audio signal and described is extracted respectively
The vocal print feature vector of first linear echo component;
Calculate the vocal print feature arrow of the first linear echo component described in the vocal print feature vector sum of first audio signal
Similarity between amount;
If the similarity being calculated is greater than preset threshold value, judgement is currently at double speaking state, and otherwise, judgement is worked as
Preceding be in singly says state.
In some alternative embodiments of the invention, the audio receiver in the audio receiver array is according to endfire array
Mode arranges placement.
The echo trap that some embodiments of the present invention provide, comprising:
First echo suppressing unit obtains first for inhibiting to the linear echo component in the first audio signal
Echo suppressed signal;
Second echo suppressing unit is used for according to first echo suppressed signal, to non-thread in the second audio signal
Property echo component is inhibited, and the second echo suppressed signal is obtained;Wherein, first audio signal and second audio letter
It number is that audio receiver in audio receiver array collects;
Third echo suppressing unit, for in second echo suppressed signal linear echo component and remain it is non-thread
Property echo component is inhibited.
In some alternative embodiments of the invention, the first echo suppressing unit, comprising:
First sef-adapting filter obtains the first linear echo component for being filtered to far-end audio signal;
First echo suppressing module is used for according to first linear echo component, in first audio signal
Linear echo component is inhibited, and the first echo suppressed signal is obtained.
In some alternative embodiments of the invention, first sef-adapting filter is also used to: singly saying shape being currently at
When state, coefficient vector is updated.
In some alternative embodiments of the invention, the second echo suppressing unit, comprising:
Second sef-adapting filter, for being filtered to obtain the first nonlinear echoes to first echo suppressed signal
Component;
Second echo suppressing module is used for according to first non linear echo components, in second audio signal
Non linear echo components inhibited, obtain the second echo suppressed signal.
In some alternative embodiments of the invention, second sef-adapting filter is also used to: singly saying shape being currently at
When state and the mean power of first echo suppressed signal are greater than the first predetermined threshold value, or state is singly said being currently at
And first linear echo component mean power be greater than the second predetermined threshold value when, coefficient vector is updated, described first
Linear echo component is filtered to obtain using the first sef-adapting filter to far-end audio signal.
In some alternative embodiments of the invention, the third echo suppressing unit, comprising:
Third sef-adapting filter obtains the second linear echo component for being filtered to far-end audio signal;
Third echo suppressing module is used for according to second linear echo component, to second echo suppressed signal
In linear echo component inhibited, obtain third echo suppressed signal;
Nonlinear echoes suppressor is used for according to second echo suppressed signal and the third echo suppressed signal,
Residual non linear echo components in the third echo suppressed signal are inhibited.
In some alternative embodiments of the invention, the third sef-adapting filter is also used to: singly saying shape being currently at
When state, coefficient vector is updated.
In some alternative embodiments of the invention, further includes: judging unit, for according to first audio signal and the
One linear echo component extracts the vocal print feature vector of first audio signal and first linear echo component respectively;
For calculate the first linear echo component described in the vocal print feature vector sum of first audio signal vocal print feature vector it
Between similarity;And if the similarity being calculated is greater than preset threshold value, judgement is currently at double speaking state, no
Then, judgement, which is currently at, singly says state.
In the above embodiment of the present invention, the first audio for being collected for the audio receiver in audio receiver array
Signal and the second audio signal obtain the suppression of the first echo by inhibiting to the linear echo component in the first audio signal
Signal processed inhibits the non linear echo components in the second audio signal further according to the first echo suppressed signal, obtains
Two echo suppressed signals finally carry out the linear echo component in the second echo suppressed signal with residual non linear echo components
Inhibit, realizes to inhibition linear and nonlinear acoustics echo in audio signal, therefore provided through the embodiment of the present invention
The technical solution of echo suppressing can effectively inhibit linear and nonlinear acoustic echo.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly introduced, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this
For the those of ordinary skill in field, without any creative labor, it can also be obtained according to these attached drawings
His attached drawing.
Fig. 1 is a kind of structural schematic diagram of single microphone nonlinear acoustics echo canceller in the prior art;
Fig. 2 is the exemplary construction schematic diagram of echo suppression techniques scheme provided by one embodiment of the invention;
Fig. 3 is the flow diagram for the echo suppressing method that one embodiment of the present of invention provides;
Fig. 4 is that the exemplary construction for the echo suppression techniques scheme using controller that one embodiment of the present of invention provides is shown
It is intended to;
Fig. 5 is the structural schematic diagram for the echo trap that one embodiment of the present of invention provides.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with attached drawing to the present invention make into
It is described in detail to one step, it is clear that the described embodiments are only some of the embodiments of the present invention, rather than whole implementation
Example.Based on the embodiments of the present invention, obtained by those of ordinary skill in the art without making creative efforts
All other embodiment, shall fall within the protection scope of the present invention.
In order to solve the mono echo suppressor realized in the prior art based on linear adaptive filter, Wu Fayou
Effect, which handles mobile phone and tablet computer etc., has acoustic echo of the equipment of communication function under handsfree talk mode non-linear
The problem of characteristic, although proposing the nonlinear adaptable filters such as artificial neural network in existing some technologies
Realize monophonic acoustics echo suppressing, however, the algorithm of these nonlinear adaptable filters is complicated, calculation amount is very big, therefore
It is usually unable on current commercial digital signal processor (Digital Signal Processing, DSP) chip enough very well
Realization.
Fig. 1 shows a kind of single microphone nonlinear acoustics echo canceller (Non-Linear in the prior art
Acoustic Echo Cancellation, NLAEC) structural schematic diagram.
Single microphone NLAEC is otherwise known as the nonlinear acoustics echo killer of post filtering, as shown in Figure 1, wrapping in NLAEC
Limit for length's unit impulse response (Finite Impulse Response, FIR) linear adaptive filter 101 and non-linear is included
Echo killer (Non-Linear Echo Suppressor, NLES) 102.As shown in Figure 1, the loudspeaker in sound system
104 will be by external environment in wheat in the sound after receiving far-end audio signal and when playing corresponding sound, played
Linear and nonlinear acoustic echo is generated at gram wind 103, is influenced by acoustic echo, microphone 103 will pick up
Response plays the acoustic echo signal of sound with loudspeaker 104.In NLAEC, the acoustic echo that is picked up for microphone 103
Signal estimates the linear component in acoustic echo signal using FIR linear adaptive filter 101 first, then by its from
It is subtracted in the signal that microphone 103 receives;It reapplies NLES 102 and inhibits remaining nonlinear component in acoustic echo signal,
To achieve the purpose that elimination and inhibit acoustic echo.
However when nonlinear residual echo component is larger, nonlinear echoes press down in single microphone NLAEC as shown in Figure 1
Device (NLES 102) processed will inevitably inhibit and damage near-end voice signals when inhibiting the nonlinear residual echo,
Especially when the amplitude of nonlinear echoes is suitable with near-end voice signals amplitude, nonlinear echoes suppressor NLES is to proximal end language
The loss of sound signal is very big.
It can see by above-mentioned analysis, the single microphone NLAEC in the prior art handled with post filtering is not
A kind of technical solution of ideal echo suppressing, and the nonlinear adaptable filters such as artificial neural network are due to calculating
Complexity it is high and be not easy to Project Realization, preferably can be realized there is presently no one kind and effectively inhibit linear and nonlinear sound
Learn the echo suppression techniques of echo.
In order to overcome the defect of above-mentioned existing echo suppression techniques scheme, provide one kind can effectively inhibit it is linear and non-
The echo suppression techniques scheme of linear acoustic echo, the present invention propose a kind of technical solution of echo suppressing.Implement in the present invention
In example, multipath audio signal is received by the audio receiver in audio receiver array, by audio signal wherein all the way
The inhibition for carrying out linear echo component carries out the inhibition of non linear echo components according to suppression result to another way audio signal,
And further inhibit linear echo component and non linear echo components, so that realizing has linear and nonlinear acoustic echo
Effect inhibits.
For example, to include that two audio receivers (being assumed to be two microphones A and B) are in audio receiver array
, in echo suppression techniques provided by the embodiment of the present invention, a FIR adaptive-filtering is introduced on the branch road of microphone A
Device constructs a linear acoustic echo Canceller (Linear Acoustic Echo Cancellation, LAEC), exports
It is then the non linear echo components in echo-signal;The non linear echo components are fed into second FIR sef-adapting filter,
So that being directed toward excipient device (Adaptive based on adaptive zero airspace constituted second FIR sef-adapting filter and microphone B
Notch Filter, ANF) it is gathered in inhibition non linear echo components;And it remains remaining linear echo component and remains on a small quantity non-
Linear echo component is respectively by the FIR in subsequent traditional nonlinear acoustics echo canceller NLAEC of the branch road of microphone B
Sef-adapting filter and nonlinear echoes suppressor (NLES) are handled.It can be seen that echo provided by the embodiment of the present invention
Nonlinear component is weaker in the input echo-signal of NLAEC in suppression technology, so that carrying out the NLES of subsequent processing to close
The decaying of end voice signal greatly reduces, therefore echo suppression techniques provided by the embodiment of the present invention can not only effectively press down
Producing linear and nonlinear acoustics echo, while NLAEC loss caused by near-end voice signals can also be reduced.
Fig. 2 is exemplarily illustrated showing for the exemplary construction of echo suppression techniques scheme provided by one embodiment of the invention
It is intended to.It as illustrated in FIG. 2, include that linear acoustic returns in the exemplary construction of echo suppressing provided by one embodiment of the invention
Excipient device (ANF) 202, nonlinear acoustics echo canceller (NLAEC) are directed toward in wave Canceller (LAEC) 201, adaptive zero airspace
203.It wherein, include a FIR sef-adapting filter (being denoted as AF3) in linear acoustic echo Canceller (LAEC) 201, adaptively
It includes that a time delay module (being denoted as D) and a FIR sef-adapting filter (are denoted as that zero airspace, which is directed toward in excipient device (ANF) 202,
It AF2), include that a FIR sef-adapting filter (being denoted as AF1) and one are non-in nonlinear acoustics echo canceller (NLAEC) 203
Linear echo suppressor (NLES).
It should be pointed out that FIR sef-adapting filter typically refers to the sef-adapting filter using FIR structure, adaptively
Filter is usually to be estimated as foundation with the statistical property of input and output signal, and algorithm is taken automatically to adjust filter system
Number, reaches the filter of optimum filtering characteristic, can specifically pass through the side of software mode, hardware mode or software and hardware combining
Formula is realized.The application does not limit the specific implementation of FIR sef-adapting filter.
As shown in Fig. 2, loudspeaker 204 generates echo-signal when playing sound based on the far-end audio signal received,
LAEC 201 is located at 205 place branch road of microphone, and echo-signal received by microphone 205 passes through in LAEC 201
After AF3 processing, the non linear echo components exported in echo-signal are arrived positioned at microphone 206 place branch road based on AF2 structure
At ANF 202, to be gathered in inhibition non linear echo components;Wherein, since ANF 202 is used for according to LAEC 201 to Mike
The echo suppressing result of echo-signal received by wind 205 to carry out echo to echo-signal received by microphone 206
Inhibit, therefore time delay module D is used to carry out delay process to echo-signal received by microphone 206, when meeting
Between on causality;Remaining all linear echo components and a small amount of residual nonlinear echoes are remained in the output signal of ANF 202
Component is then respectively via handling in the subsequent tradition NLAEC 203 for being located at 206 roads of microphone, so that realization is to line
The inhibition of property and non linear echo components.
The exemplary construction of echo suppression techniques scheme provided by one embodiment of the invention as illustrated in FIG. 2, can also
With referred to as dual microphone nonlinear acoustics echo canceller (Dual Microphone Non-Linear Acoustic Echo
Cancellation, DMNLAEC).
The embodiment of the present invention is described in detail with reference to the accompanying drawing.
Fig. 3 shows the flow diagram of echo suppressing method provided in an embodiment of the present invention.Optionally, which can answer
It is realized in exemplary construction as shown in Figure 2 or in exemplary construction as shown in Figure 2.
As shown in figure 3, the process of echo suppressing method provided in an embodiment of the present invention includes the following steps:
Step 301: the linear echo component in the first audio signal being inhibited, the first echo suppressed signal is obtained;
Step 302: according to the first echo suppressed signal, the non linear echo components in the second audio signal being pressed down
System, obtains the second echo suppressed signal;
Wherein, the first audio signal and the second audio signal are that the audio receiver in audio receiver array collects
's;
Step 303: the linear echo component in the second echo suppressed signal is pressed down with residual non linear echo components
System.
In some embodiments of the invention, the audio receiver in audio receiver array can be two or
It is multiple.Optionally, the first audio receiver in audio receiver array receives the first audio signal, in audio receiver array
The second audio receiver receive the second audio signal.
Optionally, audio receiver array can specifically arrange placement according to endfire array mode.
Such as when some embodiments of the present invention are applied in a mobile communication device, since mobile communication equipment is limited to
The reason in space, often can only spacing place 2 audio receivers (such as microphone) shortlyer, its spacing is usually 5 to 20
Between millimeter, optionally, this 2 microphones with short spacing are placed in a device according to end-fire mode.
In some embodiments of the invention, in step 301 to the linear echo component in the first audio signal into
Row inhibits, and obtains the process of the first echo suppressed signal, can be by using the first sef-adapting filter (than as shown in Figure 2
AF3) far-end audio signal is filtered to obtain the first linear echo component;Further according to the first linear echo component, to
Linear echo component in one audio signal is inhibited, and the first echo suppressed signal is obtained.
It in some embodiments of the invention, can also be further to the if judgement is currently at and singly says state
The coefficient vector of one sef-adapting filter is updated.
Wherein, singly saying that state can be indicates in the communication equipment for including audio receiver and audio player, only
There is audio player to play the state of sound based on far-end audio signal, correspondingly, double speaking state is indicated in audio player
While playing sound based on far-end audio signal, audio receiver receives the near-end audio signal that near-end user sounding obtains
State.
In some embodiments of the invention, the second audio is believed according to the first echo suppressed signal in step 302
Non linear echo components in number are inhibited, and the second echo suppressed signal is obtained, and can be by using the second adaptive filter
Wave device (such as AF2 shown in Fig. 2) is filtered the first echo suppressed signal to obtain the first non linear echo components;Further according to
First non linear echo components inhibit the non linear echo components in the second audio signal, obtain the second echo suppressing
Signal.
Wherein, it is based on step 301 obtained first time due to carrying out echo suppressing to the second audio signal in step 302
Wave inhibits signal, therefore, in order to meet the causality in timing, carries out the second audio signal of echo suppressing in step 302
It is to obtain the second audio signal received by audio receiver progress delay process, delay process does not lose the second sound
Frequency signal.For example a kind of schematical expression may is that and assume that delay parameter is D, the second audio that audio receiver receives
Signal is d1(n), then d can be expressed as by carrying out the second audio signal of echo suppressing in step 3021(n-D)。
In some embodiments of the invention, if judgement, which is currently at, singly says state and the first echo suppressed signal
Mean power is greater than the first predetermined threshold value, or judgement is currently at the average function for singly saying state and the first linear echo component
Rate be greater than the second predetermined threshold value, then can also the coefficient vector further to the second sef-adapting filter be updated, it is described
First linear echo component is filtered to obtain using the first sef-adapting filter to far-end audio signal.
In some embodiments of the invention, to the linear echo in the second echo suppressed signal point in step 303
Amount and residual non linear echo components are inhibited, and can be (such as shown in Fig. 2 by using third sef-adapting filter
AF1) far-end audio signal is filtered to obtain the second linear echo component;Further according to the second linear echo component, to second
Linear echo component in echo suppressed signal is inhibited, and third echo suppressed signal is obtained;Further according to the second echo suppressing
Signal and third echo suppressed signal, using nonlinear echoes suppressor to the residual in third echo suppressed signal non-linear time
Wave component is inhibited.
It in some embodiments of the invention, can also be further to the if judgement is currently at and singly says state
The coefficient vector of three sef-adapting filters is updated.
It, below will knot in order to more clearly describe echo suppressing method provided by the embodiment of the present invention as shown in Figure 3
Close showing for the dual microphone nonlinear acoustics echo canceller DMNLAEC that one embodiment of the present of invention as shown in Figure 2 provides
Example structure is specifically described the technical solution of echo suppressing provided by the embodiment of the present invention.
Based on exemplary construction as shown in Figure 2, in some embodiments of the invention, audio receiver array can be with
It is microphone array, specifically can be in the microphone array two of microphone 205 and microphone 206 are located at different location
Audio receiver.
Wherein, the audio signal received with microphone 205 uses d for the first audio signal2(n) it indicates;With microphone
206 audio signals received are the second audio signal, use d1(n) it indicates;Far-end audio signal is received by loudspeaker 204
The audio signal from remote subscriber, useIt indicates;Received far-end audio signal is played in loudspeaker 204When, the first audio signal d2(n) in by include microphone 205 receive by loudspeaker 204 play far-end audio letter
NumberThe echo-signal of generation will include being played by loudspeaker 204 of receiving of microphone 206 in the second audio signal
Far-end audio signalThe echo-signal of generation;Wherein, n indicates the n moment, and n is positive integer.
In some embodiments of the invention, Mike is obtained by linear acoustic echo Canceller (LAEC) 201 first
The first audio signal d that wind 205 receives2(n), wherein the first audio signal d2(n) can be interpreted as in step 301
One audio signal.
Further, linear acoustic echo Canceller (LAEC) 201 can be used for executing described in step 301, right
Linear echo component in first audio signal is inhibited, and the process of the first echo suppressed signal is obtained, which can also be with
It is interpreted as through linear acoustic echo Canceller (LAEC) 201 from the first audio signal d2(n) it is extracted in the echo-signal in non-
Linear echo component, obtained result are the first echo suppressed signal.Specifically, linear acoustic echo Canceller (LAEC)
201 can pass through FIR sef-adapting filter (AF3) Lai Zhihang step 301.
Wherein, FIR sef-adapting filter (AF3) can be based on from far-end audio signalFirst, in accordance with following public affairs
Formula is to far-end audio signalIt is filtered, to estimate the first linear echo component:
Wherein, y3(n) the first linear echo component is indicated,It is FIR sef-adapting filter (AF3) at the n moment
Coefficient vector,L3Indicate FIR sef-adapting filter
(AF3) dimension possessed by coefficient vector, n, L3For positive integer;For the vector at far-end audio signal n moment,T is the transposed operator in vector operation.
Based on obtained first linear echo component y3It (n), can in linear acoustic echo canceller (LAEC) 201
Further to the first audio signal d2(n) pass through formula e3(n)=d2(n)-y3(n), to linear in the first audio signal
Echo component is inhibited, and the first echo suppressed signal, e are obtained3(n) the first echo suppressed signal is indicated.
It should be pointed out that the first echo suppressed signal e obtained by the above process3(n), in the first audio signal d2
(n) not including in will be only comprising non-linear time in echo-signal under conditions of near-end voice signals (or also known as singly say state)
Wave component, and in the first audio signal d2(n) including in then will packet under conditions of near-end speech (or also known as double speaking state)
Contain the non linear echo components and the near-end voice signals in echo-signal.
Further, be currently at singly say state when, can be to the coefficient vector of FIR sef-adapting filter (AF3)It is updated.Specifically, the coefficient vector of FIR sef-adapting filter (AF3)Update (or be referred to as
Study) can be used such as normalization minimum mean-square calculation (Normalized Least Mean Square algorithm,
) or affine projection algorithm (Affine Projection Algorithm, APA) or recursive least squares NLMS
Adaptive learnings algorithms such as (Recursive Least Square, RLS) carries out, wherein in the case where singly saying state,
The coefficient study of AF3, which updates, can be lasting progress, and in the case where double say, the coefficient study of AF3 updates and can stop.
As an example, in some embodiments of the invention, it is based on the first echo suppressed signal e3(n) and it is remote
Hold audio signalPass through NLMS algorithm, AF3 coefficient vectorIt can be updated according to following formula:
Wherein,Indicate the coefficient vector of FIR sef-adapting filter (AF3) at the n moment,Certainly for FIR
Coefficient vector of the adaptive filter (AF3) at the n+1 moment;0<μ3< 1 is update compensating parameter, δ3> 0 is regularization factor parameter;
The coefficient vector that (i) therein formula is currently at FIR sef-adapting filter (AF3) when singly saying state by judgement, which is updated, to be made
Formula;(ii) formula indicates the coefficient vector of the FIR sef-adapting filter (AF3) when judgement is currently at double speaking state not more
Newly.
Further, the exemplary construction based on one embodiment of the present of invention shown in Fig. 2, of the invention some specific
In embodiment, the above-mentioned obtained first echo suppressed signal e of FIR sef-adapting filter (AF3)3(n), it can further feed
It is directed toward in excipient device (ANF) 202 and is handled to adaptive zero airspace.
The the first audio signal d received due to microphone 2052(n) non linear echo components in connect with microphone 206
The the second audio signal d received1(n) there is high correlations between the non linear echo components in, to filter FIR adaptively
Zero airspace that excipient device (ANF) is directed toward in adaptive zero airspace that wave device (AF2) and microphone 206 are constituted, which is directed toward, is gathered in the second sound
Frequency signal d1(n) non linear echo components in, a kind of optional processing mode are exactly to include what is obtained in step 302
First audio signal d2(n) the first echo suppressed signal e of the non linear echo components in3(n) it is fed to FIR adaptive-filtering
As input in device (AF2), and the update of the coefficient vector of FIR sef-adapting filter (AF2) then singly says state being currently at
And non linear echo components biggish condition carries out when setting up, wherein non linear echo components are larger can refer to it is non-linear
Energy possessed by echo component is higher than preset threshold value.
It should be pointed out that in some embodiments of the invention, the distance between microphone 205 and microphone 206 compared with
Short, the two-microphone array of short spacing both can be used side when placing and penetrate mode (broad-side), and end-fire mode can also be used
(end-fire), and when there are proximal end target audio signal, FIR sef-adapting filter (AF2) and microphone 206 constitute from
Zero airspace for adapting to zero airspace direction excipient device (ANF), which is directed toward, also will affect the near-end voice signals being directed toward from zero airspace.
In view of the position of sound production mode of dual microphone will affect the spy that excipient device (ANF) is directed toward in adaptive zero airspace
Property, excipient device more adaptive than the dual microphone that end-fire is placed zero is directed toward in view of adaptive zero airspace of dual microphone that placement is penetrated on side
Excipient device (ANF) is directed toward in airspace, and there is zero relatively wide airspace to be directed toward, and thus be easier to cause to decay to near-end voice signals,
Therefore a kind of optionally modes of emplacement is that dual microphone (microphone 205 and microphone 206 as shown in Figure 2) uses end-fire side
Formula is placed, to be avoided as much as or reduce the decaying caused by near-end audio signal.Wherein, microphone 205 is located at from audio
Player (than loudspeaker as shown in Figure 2 (Speaker) 204) closer position, is arranged linear acoustic echo on its branch road
Canceller (LAEC) 201, and microphone 206 is located at from 204 remote position of loudspeaker, and it is empty that adaptive zero is arranged on its branch road
It is directed toward excipient device (ANF) and nonlinear acoustics echo canceller (NLAEC) in domain.
Specifically, second that the available microphone 206 of excipient device (ANF) one side receives is directed toward in adaptive zero airspace
Audio signal d1(n), the second audio signal d1(n) the second audio signal that can be interpreted as in step 301;On the other hand, also
It can be based on the first echo suppressed signal e received3(n), it executes described in step 302 and is believed according to the first echo suppressing
Number, the non linear echo components in the second audio signal are inhibited, the second echo suppressed signal is obtained.
Further, adaptive zero airspace is directed toward excipient device (ANF) 202 and can adaptively be filtered by FIR wherein included
The process of wave device (AF2) Lai Zhihang step 302.Specifically, FIR sef-adapting filter (AF2) can be according to the received
One echo suppressed signalFirst, in accordance with following formula to the first echo suppressed signalIt is filtered, to estimate
Count the first non linear echo components:
Wherein, y2(n) the first non linear echo components are indicated,It is FIR sef-adapting filter (AF2) at the n moment
Coefficient vector,L2For FIR sef-adapting filter
(AF2) dimension of coefficient vector, n, L2For positive integer;For the first echo suppressed signal of n moment;
Based on obtained first non linear echo components y2(n), excipient device (ANF) 202 is directed toward in adaptive zero airspace
In, it can be further to the second audio signal d after delay process1(n-D) pass through formula e2(n)=d1(n-D)-y2(n),
Non linear echo components in second audio signal are inhibited, the second echo suppressed signal is obtained, uses e2(n) it indicates.
Wherein, d1(n-D) the second audio signal d is indicated1(n) according to the signal after delay parameter D delay process, due to certainly
It adapts to zero airspace and is directed toward the second audio signal d that excipient device (ANF) is got at the n moment1(n), echo suppression is not immediately engaged in
The process of system, therefore adaptive zero airspace direction excipient device (ANF) will be to the second audio signal d1(n) delay process (ratio is carried out
D delay component as shown in Figure 2) to meet causality, so that the second audio signal after being delayed, uses d1(n-D) table
Show, normally, the value of D can be according to D=round (L2/ 2) it is calculated.
Specifically, the coefficient vector of FIR sef-adapting filter (AF2) singly says state and nonlinear echoes being currently at
Energy possessed by component is updated when being higher than preset threshold value, wherein in view of the function of the linear component in acoustic echo
In the power of its corresponding nonlinear component, energy possessed by non linear echo components has rate ratio higher than preset threshold value
The mean power that body can show as the first linear echo component is greater than the second predetermined threshold value or the first echo suppressed signal
Mean power be greater than first predetermined threshold value etc..Further, if judgement, which is currently at, is singly said state and first echo
The mean power of signal is inhibited to be greater than the first predetermined threshold value, or judgement is currently at and singly says state and the first linear echo point
The mean power of amount is greater than the second predetermined threshold value, then can carry out more to the coefficient vector of FIR sef-adapting filter (AF2)
Newly.
It should be pointed out that determining that the mode of this larger condition of non linear echo components includes but is not limited to that the present invention is real
Aforesaid way provided by example is applied, for example can also be based on any object for being able to reflect nonlinear echoes energy or watt level
Reason amount and a combination thereof etc..
Specifically, similarly with the update of the coefficient vector of aforementioned FIR sef-adapting filter (AF3), it can be used such as
The adaptive learnings such as NLMS or APA algorithm is updated the coefficient vector of FIR sef-adapting filter (AF2).
As an example, in some embodiments of the invention, it is based on the first echo suppressed signal e3(n) and the
Two echo suppressed signal e2(n), by NLMS algorithm, AF2 coefficient vector can be updated according to following formula:
Wherein,Coefficient vector for FIR sef-adapting filter (AF2) at the n moment,Certainly for FIR
Coefficient vector of the adaptive filter (AF2) at the n+1 moment;0<μ2< 1 is update compensating parameter, δ2> 0 is regularization factor parameter;
(i) formula be judgement be currently at singly say state and further judge non linear echo components it is larger (such as be specially first linearly
The mean power of echo component is greater than the second predetermined threshold value or the mean power of the first echo suppressed signal is greater than first in advance
Gating limit value etc.) when FIR sef-adapting filter (AF2) the formula that is updated of coefficient vector;(ii) formula expression is being unsatisfactory for
(i) coefficient vector of FIR sef-adapting filter (AF2) does not update when formula condition.
It should be pointed out that the second audio signal d that microphone 206 receives1(n) refer to by above-mentioned adaptive zero airspace
To after excipient device (ANF) processing, the second echo suppressed signal e of output2(n) include prevailing linear echo point in
Amount and a small amount of residual nonlinear echoes, therefore to e2(n) signal can directly apply traditional nonlinear acoustics echo canceller
NLAEC is handled, and can obtain satisfactory result.
Wherein, NLAEC can actually be that LAEC and nonlinear echoes suppressor (NLES) cascade are constituted, using with L1
The FIR sef-adapting filter (AF1) of coefficient vector is tieed up to estimate e2(n) linear echo component in, and by it from e2(n) subtract in
It goes, inhibits prevailing linear echo component to reach, reuse nonlinear echoes suppressor (NLES) suppression of postposition
Make remaining non linear echo components.
Based on the exemplary construction of one embodiment of the present of invention shown in Fig. 2, in some embodiments of the invention,
The above-mentioned obtained second echo suppressed signal e of FIR sef-adapting filter (AF2)2(n), nonlinear acoustics echo can be fed to
Subsequent processing is participated in Canceller (NLAEC) 203.
Specifically, nonlinear acoustics echo canceller (NLAEC) 203 can be used for executing in step 303 and press down to the second echo
The process that linear echo component and residual non linear echo components in signal processed are inhibited.Wherein, nonlinear acoustics echo
Canceller (NLAEC) 203 can be filtered far-end audio signal by FIR sef-adapting filter (AF1) wherein included
To the second linear echo component, according to the second linear echo component, to the linear echo in second echo suppressed signal point
Amount is inhibited, and third echo suppressed signal is obtained;Further according to second echo suppressed signal and the third echo suppressing
Signal, by nonlinear echoes suppressor (NLES) wherein included to the residual nonlinear echoes in third echo suppressed signal
Component is inhibited.
Specifically, the FIR sef-adapting filter (AF1) in nonlinear acoustics echo canceller (NLAEC) 203 can be root
According to the far-end audio signal receivedFirst, in accordance with following formula to far-end audio signalIt is filtered, to estimate
Count the second linear echo component:
Wherein, y1(n) the second linear echo component is indicated,It is at the n moment for FIR sef-adapting filter (AF1)
Number vector,L1It indicates FIR sef-adapting filter (AF1)
The dimension that coefficient vector has, n, L1For positive integer;Indicate the far-end audio signal phasor at n moment,
Based on the second obtained linear echo component y1It (n), can in nonlinear acoustics echo canceller (NLAEC) 203
Further from the second echo suppressed signal e2(n) pass through formula e in1(n)=e2(n)-y1(n), the second echo suppressing is believed
Linear echo component in number is inhibited, and is obtained third echo suppressed signal, is used e1(n) it indicates.
Specifically, with the coefficient vector of aforementioned FIR sef-adapting filter (AF3) and FIR sef-adapting filter (AF2)
It updates similarly, the coefficient vector of FIR sef-adapting filter (AF1)Can also when judgement is currently at and singly says state into
Row updates, for example is updated using the adaptive learnings such as NLMS or APA or RLS algorithm.
As an example, in some embodiments of the invention, it is based on third echo suppressed signal e1(n) and it is remote
Hold audio signalBy NLMS algorithm, AF1 coefficient vector can be updated according to following formula:
Wherein,Coefficient vector for FIR sef-adapting filter (AF1) at the n moment,Certainly for FIR
Coefficient vector of the adaptive filter (AF1) at the n+1 moment;0<μ2< 1 is update compensating parameter, δ2> 0 is regularization factor parameter;
(i) formula is that judgement is currently at the formula that the coefficient vector of FIR sef-adapting filter (AF1) when singly saying state is updated;
(ii) formula indicates that the coefficient vector of FIR sef-adapting filter (AF1) does not update when judgement is currently at double speaking state.
Further, third echo suppressed signal e1(n) the residual non linear echo components in can pass through nonlinear acoustics
Nonlinear echoes suppressor (NLES) in echo canceller (NLAEC) 203 is further processed.Due to remaining at this time non-linear
Echo component is weaker, then the inhibition yield value of NLES is usually larger (no more than 1 but close to 1), so if in third echo
Inhibit signal e1It (n) include proximal end target audio signal in, then NLES will not cause decaying to it or decaying is less.Specifically
Ground, NLES can be according to the second echo suppressed signal of its input signal e2(n) and third echo suppressed signal e1(n), first in frequency
A frequency spectrum modified gain is generated in domain, and third echo suppressed signal e is corrected with the gain1(n) frequency spectrum, to reach suppression
Third echo suppressed signal e processed1(n) purpose of residual nonlinear echoes in.
Optionally, the nonlinear echoes suppressor (NLES) in nonlinear acoustics echo canceller (NLAEC) 203 can root
According to minimum mean square error criterion, according to the second echo suppressed signal e of input2(n) and third echo suppressed signal e1(n), first
Frequency spectrum amendment is generated in frequency domain according to following formula and inhibits gain:
Wherein, G (m, k) indicates that NLES frequency spectrum amendment generated inhibits gain,Indicate e1(n) and e2(n)
Between crosspower spectrum,Indicate signal e1(n) auto-power spectrum is calculated by the following formula to obtain respectively:
Wherein, E2(m, k) is the second echo suppressed signal e2(n) Short Time Fourier Transform of than the m-th data block, E1(m,
It k) is third echo suppressed signal e1(n) Short Time Fourier Transform of than the m-th data block, k are frequency point index, and conj { } is multiple
Adjoint operator, λ are smoothing factor constant, and 0 < λ < 1, usual λ can be with value in 0.925 to 0.999 range.
In turn, gain G (m, k) can be inhibited based on frequency spectrum generated amendment, corrects third echo according to following formula
Inhibit signal e1(n) E1(m,k);Y (m, k)=E1(m, k) G (m, k) carries out in Fu in short-term obtained Y (m, k) again
The output signal y (n) of nonlinear echoes suppressor (NLES) can be obtained in leaf inverse transformation, it can be seen that received
When in two audio signals including echo-signal and near-end audio signal, which is to inhibit linearity and non-linearity echo component
For being output to the proximal end target audio signal of remote subscriber.
Further, in some embodiments of the invention, it is directed in the above process, linear acoustic echo is offset
The FIR that FIR sef-adapting filter (AF1), adaptive zero airspace in device (LAEC) 201 are directed toward in excipient device (ANF) 202 is adaptive
Answer the FIR sef-adapting filter (AF1) in filter (AF2) and nonlinear acoustics echo canceller (NLAEC) 203 is
The update of number vector can realize control by controller.
Specifically, in some embodiments of the invention, controller can be presently at by judgement double says shape
State, which is in, singly says state, to realize to FIR sef-adapting filter (AF3), FIR sef-adapting filter (AF2) and FIR
The control that the coefficient vector of sef-adapting filter (AF1) updates.
Optionally, in some embodiments of the invention, controller can judge in the following ways it is current whether
In singly saying state:
According to the first audio signal and the first linear echo component, the first audio signal and the first linear echo are extracted respectively
The vocal print feature vector of component;Calculate the vocal print feature of the first linear echo component of vocal print feature vector sum of the first audio signal
Similarity between vector;If the similarity being calculated is greater than preset threshold value, judgement is currently at double speaking state, no
Then, judgement, which is currently at, singly says state.Wherein it is possible to similarity calculation be carried out using many algorithms, for example, the first sound can be calculated
Euclidean distance between the vocal print feature vector of the first linear echo component of vocal print feature vector sum of frequency signal, uses Euclidean distance
The similarity of characterization between the two.
Based on exemplary construction illustrated in fig. 2, Fig. 4 shows the echo suppression techniques of some embodiments of the present invention offer
The schematic diagram updated in scheme using controller control FIR adaptive filter coefficient vector.Wherein, Fig. 4 it is shown in Fig. 2 go out
Controller 401 is further comprised in structure, which is used to control in AF3, ANF 202 in LAEC 201
The update of the coefficient vector of AF2 and the AF1 in NLAEC 203.
That as shown in Figure 4 includes the exemplary construction of controller 401, the available first audio signal d of controller 4012
(n) and the first linear echo component y3(n);Controller 401 again by extracting the first audio signal d respectively2(n) and first is linear
Echo component y3(n) vocal print feature vector further calculates the first audio signal d2(n) vocal print feature vector sum and
One linear echo component y3(n) similarity between vocal print feature vector;If the similarity being calculated is greater than preset
Threshold value, the then judgement of controller 401 are currently at double speaking state, and otherwise, the judgement of controller 401, which is currently at, singly says state.
Further, controller 401 is being judged to be currently at double speaking state or judge to be currently at singly to say state
Afterwards, judging result can be fed in the AF2 in AF3, ANF 202 in the LAEC 201 and AF1 in NLAEC 203,
As controller 401 is directed toward shown in external arrow in Fig. 4.
And then for the AF1 in the AF3 and NLAEC 203 in LAEC 201, can be fed according to controller 401
Current state carries out the update of coefficient vector when being currently at and singly saying state, stops coefficient when being currently at double speaking state
The update of vector, specific renewal process can be found in described previously.
And for the AF2 in ANF 202, since it only singly says state and non linear echo components are larger being currently at
Condition under, just carry out coefficient vector update, therefore, for the AF2 in ANF 202, controller 401 is also used to determine
Whether non linear echo components are larger and to 202 feedback and judge result of ANF.Wherein, as it was noted above, non linear echo components
The larger energy that can be understood as non linear echo components is higher than preset threshold value.Specifically, it is mentioned in the embodiment of the present invention
In the echo suppression techniques scheme of confession, the energy of non linear echo components, which is higher than preset threshold value, can show as the first echo
Inhibit signal e3(n) mean power is greater than the first predetermined threshold value, or can be First Line echo component y3(n) flat
Equal power is greater than the second predetermined threshold value.
In turn, controller 401 can be according to the first linear echo component y3(n) mean power determines nonlinear echoes
It is whether larger, for example illustratively indicate to determine as a result, sentencing according to example as shown below with nonlinear echoes size state
It is fixed:
Wherein, Py3It (n) is the first linear echo component y3(n) mean power can be calculated: P according to following formulay3
(n)=α Py3(n-1)+(1-α)·{y3(n)}2;α is smoothing factor constant, and 0 < α < 1, usual value is 0.925 to 0.999
In range;ThresholdLEIndicate the second predetermined threshold value.
Alternatively, controller 401 can also be according to the first echo suppressed signal e3(n) mean power determines non-linear time
Whether wave larger, for example, illustratively with nonlinear echoes size state indicate determine as a result, according to example as shown below come
Determine:
Wherein, PNLEIt (n) is the first echo suppressed signal e3(n) mean power can be calculated according to following formula:Wherein α is smoothing factor constant, and 0 < α < 1 usually takes
Value is in 0.925 to 0.999 range;ThresholdNLEIndicate the first predetermined threshold value;(i) formula is to be currently at list in judgement
First echo suppressed signal e when saying state3(n) calculation formula of mean power, (ii) formula are to be currently at double say in judgement
First echo suppressed signal e when state3(n) calculation formula of mean power.
In turn, controller 401 can be to 202 feedback and judge of ANF as a result, AF2 root in ANF 202 after being determined
According to controller 401 feed current state and judgement nonlinear echoes size state, be currently at singly say state and
The update that coefficient vector is carried out when nonlinear echoes are larger, stops the update of coefficient vector when being unsatisfactory for above-mentioned condition, specifically
Renewal process can be found in described previously.
In conclusion in the echo suppressing method provided in embodiments of the present invention, in audio receiver array
The first audio signal and the second audio signal that audio receiver receives, by the linear echo in the first audio signal point
Amount is inhibited, and the first echo suppressed signal is obtained, further according to the first echo suppressed signal, to non-thread in the second audio signal
Property echo component inhibited, the second echo suppressed signal is obtained, finally to the linear echo in the second echo suppressed signal point
Amount and residual non linear echo components are inhibited, and are realized to inhibition linear and nonlinear acoustics echo in audio signal,
Therefore the technical solution of provided echo suppressing can effectively inhibit linear and nonlinear acoustics through the embodiment of the present invention
Echo.
In addition, the technical solution of echo suppressing provided by the embodiment of the present invention is effectively pressing down for compared with the prior art
While producing linear and nonlinear acoustics echo, caused by near-end voice signals damage very little or it is some ideally
It can achieve the effect that not damage.
For example, the diamylose of short spacing is applied to for the echo suppressing method for providing some embodiments of the present invention
During carrying out echo suppressing in gram wind array, by linear acoustic echo Canceller (LAEC) from one of microphone
Branch extracts nonlinear echoes, and the non linear echo components of the extraction are then conveyed to adaptive zero sky in another microphone branch
Domain is directed toward excipient device (ANF) and is used as reference signal, so that zero airspace of ANF, which is directed toward, is gathered in inhibition nonlinear echoes;ANF's is defeated
It is handled out using traditional nonlinear acoustics echo canceller (NLAEC), eliminates its linear echo component, while further
Inhibit remaining compared with small nonlinearity echo component.It is this by echo suppressing method application provided by some embodiments of the present invention
To the scheme for carrying out echo suppressing based on short spacing two-microphone array, it can not only effectively inhibit what microphone received
Acoustic echo (including linear and nonlinear echo) in signal can also reduce the damage to near-end speech simultaneously.
It should be understood that when, in this application, be mainly provided for the embodiments of the invention echo from time domain angle
The technical solution of inhibition is specifically described, and the technical solution specific implementation of echo suppressing provided by the embodiment of the present invention includes
But it is not limited to the time domain way of realization that the application is shown, for example can also be frequency domain, subband domain and wavelet transformed domain etc. and is real
Existing form.
Based on the same technical idea, the embodiment of the present invention also provides a kind of echo trap, which can be by soft
Part mode, hardware mode or software and hardware combining mode realize that the echo suppression that the above embodiment of the present invention provides can be performed in the device
Embodiment of the method processed.
Optionally, which can be applicable in the exemplary construction that some embodiments of the present invention as illustrated in FIG. 2 provide,
Or it can be applicable in the exemplary construction that other embodiment of the invention as illustrated in FIG. 4 provides;Alternatively, the device can pass through
The exemplary construction that some embodiments of the present invention as illustrated in FIG. 2 provide is realized, or can pass through the present invention as illustrated in FIG. 4
Other embodiment provide exemplary construction realize.
Fig. 5 shows the structural schematic diagram of the echo trap of some embodiments of the present invention offer, as shown in figure 5,
The device includes:
First echo suppressing unit 501 obtains for inhibiting to the linear echo component in the first audio signal
One echo suppressed signal;
Second echo suppressing unit 502 is used for according to first echo suppressed signal, to non-in the second audio signal
Linear echo component is inhibited, and the second echo suppressed signal is obtained;Wherein, first audio signal and second audio
Signal is that the audio receiver in audio receiver array collects;
Third echo suppressing unit 503, for the linear echo component and residual in second echo suppressed signal
Non linear echo components are inhibited.
Optionally, the first echo suppressing unit 501, can specifically include:
First sef-adapting filter 5011 obtains the first linear echo component for being filtered to far-end audio signal;
First echo suppressing module 5012 is used for according to first linear echo component, to first audio signal
In linear echo component inhibited, obtain the first echo suppressed signal.
Further, the first sef-adapting filter 5011, can be also used for be currently at singly say state when, to coefficient swear
Amount is updated.
The exemplary construction that some embodiments of the present invention gone out as shown in connection with fig. 2 provide, the first echo suppressing unit 501,
It specifically can be linear acoustic echo Canceller (LAEC) 201 illustrated in fig. 2;Or go out as shown in connection with fig. 4 of the invention
The exemplary construction that other embodiment provides, the first echo suppressing unit 501 specifically can be linear acoustic illustrated in fig. 4
Echo canceller (LAEC) 201;Therefore, the concrete property of the first echo suppressing unit 501 and function can be found in aforementioned implementation
Description in example to linear acoustic echo Canceller (LAEC) 201, will not be described in great detail herein.
Optionally, the second echo suppressing unit 502, can specifically include:
Second sef-adapting filter 5021, it is non-linear for being filtered to obtain first to first echo suppressed signal
Echo component;
Second echo suppressing module 5022, for believing second audio according to first non linear echo components
Non linear echo components in number are inhibited, and the second echo suppressed signal is obtained.
Further, the second sef-adapting filter 5021 can be also used for singly saying state and described first being currently at
When the mean power of echo suppressed signal is greater than the first predetermined threshold value, or being currently at singly say state and first linear returns
When the mean power of wave component is greater than the second predetermined threshold value, coefficient vector is updated, wherein the first linear echo component
The first sef-adapting filter is used to be filtered to obtain to far-end audio signal.
The exemplary construction that some embodiments of the present invention gone out as shown in connection with fig. 2 provide, the second echo suppressing unit 502,
It specifically can be the component part that adaptive zero airspace illustrated in fig. 2 is directed toward in excipient device (ANF) 202;Or combine Fig. 4 institute
The exemplary construction that the other embodiment of the invention shown provides, the second echo suppressing unit 502 specifically can be Fig. 4 institute
It is directed toward the component part in excipient device (ANF) 202 in adaptive zero airspace shown;Therefore, the tool of the second echo suppressing unit 502
Bulk properties and function can be found in the description for being directed toward excipient device (ANF) 202 in previous embodiment to adaptive zero airspace, herein will
It repeats no more.
Optionally, third echo suppressing unit 503, can specifically include:
Third sef-adapting filter 5031 obtains the second linear echo component for being filtered to far-end audio signal;
Third echo suppressing module 5032 is used for according to second linear echo component, to second echo suppressing
Linear echo component in signal is inhibited, and third echo suppressed signal is obtained;
Nonlinear echoes suppressor 5033, for being believed according to second echo suppressed signal and the third echo suppressing
Number, the residual non linear echo components in the third echo suppressed signal are inhibited.
Further, third sef-adapting filter 5031, can be also used for be currently at singly say state when, to coefficient swear
Amount is updated.
The exemplary construction that some embodiments of the present invention gone out as shown in connection with fig. 2 provide, third echo suppressing unit 503,
It specifically can be the component part in nonlinear acoustics echo canceller (NLAEC) 203 illustrated in fig. 2;Or combine Fig. 4 institute
The exemplary construction that the other embodiment of the invention shown provides, third echo suppressing unit 503 specifically can be Fig. 4 institute
The component part in nonlinear acoustics echo canceller (NLAEC) 203 shown;Therefore, the tool of third echo suppressing unit 503
Bulk properties and function can be found in the description in previous embodiment to nonlinear acoustics echo canceller (NLAEC) 203, herein will
It repeats no more.
Further, can also include: in the echo trap provided by some embodiments of the present invention
Judging unit 504, for according to first audio signal and the first linear echo component, extracting described the respectively
The vocal print feature vector of one audio signal and first linear echo component;And
Vocal print for calculating the first linear echo component described in the vocal print feature vector sum of first audio signal is special
Levy the similarity between vector;And
If the similarity for being calculated is greater than preset threshold value, judgement is currently at double speaking state, otherwise, sentences
Disconnected be currently at singly says state.
Correspondingly, the exemplary construction that some embodiments of the present invention gone out as shown in connection with fig. 4 provide, judging unit 504, tool
Body can be the component part in controller 401 illustrated in fig. 4;Therefore, the concrete property of judging unit 504 and function can
Referring to the description in previous embodiment to controller 401, also will not be described in great detail herein.
Optionally, the audio receiver in audio receiver array arranges placement according to endfire array mode.
For software implementation, these technologies can use module (such as program, the function etc. for realizing function described herein
Deng) realize.Software code can store in a memory cell, and be executed by processor.Memory cell can handled
It realizes in device or outside the processor.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications may be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (17)
1. a kind of echo suppressing method, which is characterized in that this method comprises:
Linear echo component in first audio signal is inhibited, the first echo suppressed signal is obtained;
According to first echo suppressed signal, the non linear echo components in the second audio signal are inhibited, obtain
Two echo suppressed signals;Wherein, first audio signal and second audio signal are the sounds in audio receiver array
Frequency receiver collects;
Linear echo component in second echo suppressed signal is inhibited with residual non linear echo components.
2. the method as described in claim 1, which is characterized in that press down to the linear echo component in the first audio signal
System, obtains the first echo suppressed signal, comprising:
Far-end audio signal is filtered using the first sef-adapting filter to obtain the first linear echo component;
According to first linear echo component, the linear echo component in first audio signal is inhibited, is obtained
First echo suppressed signal.
3. method according to claim 2, which is characterized in that further include: if judgement, which is currently at, is singly said state, to described
The coefficient vector of first sef-adapting filter is updated.
4. the method as described in claim 1, which is characterized in that according to first echo suppressed signal, believe the second audio
Non linear echo components in number are inhibited, and the second echo suppressed signal is obtained, comprising:
First echo suppressed signal is filtered to obtain the first non linear echo components using the second sef-adapting filter;
According to first non linear echo components, the non linear echo components in second audio signal are inhibited,
Obtain the second echo suppressed signal.
5. method as claimed in claim 4, which is characterized in that further include: if judgement, which is currently at, singly says state and described the
The mean power of one echo suppressed signal is greater than the first predetermined threshold value, or judgement is currently at and singly says state and first linearly
The mean power of echo component is greater than the second predetermined threshold value, then carries out more to the coefficient vector of second sef-adapting filter
Newly, first linear echo component is filtered to obtain using the first sef-adapting filter to far-end audio signal.
6. the method as described in claim 1, which is characterized in that the linear echo component in second echo suppressed signal
Inhibited with residual non linear echo components, comprising:
Far-end audio signal is filtered using third sef-adapting filter to obtain the second linear echo component;
According to second linear echo component, the linear echo component in second echo suppressed signal is inhibited,
Obtain third echo suppressed signal;
According to second echo suppressed signal and the third echo suppressed signal, using nonlinear echoes suppressor to described
Residual non linear echo components in third echo suppressed signal are inhibited.
7. method as claimed in claim 6, which is characterized in that further include: if judgement, which is currently at, is singly said state, to described
The coefficient vector of third sef-adapting filter is updated.
8. the method as described in claim 3,5 or 7, which is characterized in that current whether be in of judgement is singly said in the following ways
State:
According to first audio signal and the first linear echo component, first audio signal and described first are extracted respectively
The vocal print feature vector of linear echo component, wherein first linear echo component is to use the first sef-adapting filter pair
Far-end audio signal is filtered;
Calculate the first linear echo component described in the vocal print feature vector sum of first audio signal vocal print feature vector it
Between similarity;
If the similarity being calculated is greater than preset threshold value, judgement is currently at double speaking state, otherwise, judges current place
In singly saying state.
9. the method as described in any one of claims 1 to 7, which is characterized in that the audio in the audio receiver array
Receiver arranges placement according to endfire array mode.
10. a kind of echo trap, which is characterized in that the device includes:
First echo suppressing unit obtains the first echo for inhibiting to the linear echo component in the first audio signal
Inhibit signal;
Second echo suppressing unit is used for according to first echo suppressed signal, to non-linear time in the second audio signal
Wave component is inhibited, and the second echo suppressed signal is obtained;Wherein, first audio signal and second audio signal are
What the audio receiver in audio receiver array collected;
Third echo suppressing unit, for in second echo suppressed signal linear echo component and residual non-linear time
Wave component is inhibited.
11. device as claimed in claim 10, which is characterized in that the first echo suppressing unit, comprising:
First sef-adapting filter obtains the first linear echo component for being filtered to far-end audio signal;
First echo suppressing module is used for according to first linear echo component, to linear in first audio signal
Echo component is inhibited, and the first echo suppressed signal is obtained.
12. device as claimed in claim 11, which is characterized in that first sef-adapting filter is also used to: locating currently
When singly saying state, coefficient vector is updated.
13. device as claimed in claim 10, which is characterized in that the second echo suppressing unit, comprising:
Second sef-adapting filter obtains the first nonlinear echoes point for being filtered to first echo suppressed signal
Amount;
Second echo suppressing module is used for according to first non linear echo components, to non-in second audio signal
Linear echo component is inhibited, and the second echo suppressed signal is obtained.
14. device as claimed in claim 13, which is characterized in that second sef-adapting filter is also used to: locating currently
When singly saying that the mean power of state and first echo suppressed signal is greater than the first predetermined threshold value, or it is being currently at
When singly saying that the mean power of state and the first linear echo component is greater than the second predetermined threshold value, coefficient vector is updated,
First linear echo component is filtered to obtain using the first sef-adapting filter to far-end audio signal.
15. device as claimed in claim 10, which is characterized in that the third echo suppressing unit, comprising:
Third sef-adapting filter obtains the second linear echo component for being filtered to far-end audio signal;
Third echo suppressing module is used for according to second linear echo component, in second echo suppressed signal
Linear echo component is inhibited, and third echo suppressed signal is obtained;
Nonlinear echoes suppressor is used for according to second echo suppressed signal and the third echo suppressed signal, to institute
The residual non linear echo components stated in third echo suppressed signal are inhibited.
16. device as claimed in claim 15, which is characterized in that the third sef-adapting filter is also used to: locating currently
When singly saying state, coefficient vector is updated.
17. the device as described in claim 12,14 or 16, which is characterized in that further include:
Judging unit, for extracting first audio respectively according to first audio signal and the first linear echo component
The vocal print feature vector of signal and first linear echo component, wherein first linear echo component is to use first
Sef-adapting filter is filtered far-end audio signal;Vocal print feature for calculating first audio signal is sweared
Similarity between amount and the vocal print feature vector of first linear echo component;And if the similarity being calculated is big
In preset threshold value, then judgement is currently at double speaking state, and otherwise, judgement, which is currently at, singly says state.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610399409.4A CN107483761B (en) | 2016-06-07 | 2016-06-07 | A kind of echo suppressing method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610399409.4A CN107483761B (en) | 2016-06-07 | 2016-06-07 | A kind of echo suppressing method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107483761A CN107483761A (en) | 2017-12-15 |
CN107483761B true CN107483761B (en) | 2019-09-17 |
Family
ID=60594193
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610399409.4A Active CN107483761B (en) | 2016-06-07 | 2016-06-07 | A kind of echo suppressing method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107483761B (en) |
Families Citing this family (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107749304B (en) * | 2017-09-07 | 2021-04-06 | 电信科学技术研究院 | Method and device for continuously updating coefficient vector of finite impulse response filter |
EP3824616B1 (en) * | 2018-07-18 | 2023-09-06 | Google LLC | Echo detection |
CN109040498B (en) * | 2018-08-12 | 2022-01-07 | 瑞声科技(南京)有限公司 | Method and system for improving echo cancellation effect |
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 |
CN111292760B (en) * | 2019-05-10 | 2022-11-15 | 展讯通信(上海)有限公司 | Sounding state detection method and user equipment |
CN110956975B (en) * | 2019-12-06 | 2023-03-24 | 展讯通信(上海)有限公司 | Echo cancellation method and device |
CN112017679B (en) * | 2020-08-05 | 2024-01-26 | 海尔优家智能科技(北京)有限公司 | Method, device and equipment for updating adaptive filter coefficients |
CN112217948B (en) * | 2020-10-10 | 2022-01-18 | 北京嘀嘀无限科技发展有限公司 | Echo processing method, device, equipment and storage medium for voice call |
CN112491449B (en) * | 2020-11-04 | 2024-04-16 | 北京百度网讯科技有限公司 | Acoustic echo cancellation method, acoustic echo cancellation device, electronic equipment and storage medium |
CN113362843B (en) * | 2021-06-30 | 2023-02-17 | 北京小米移动软件有限公司 | Audio signal processing method and device |
CN113489855B (en) * | 2021-06-30 | 2024-03-19 | 北京小米移动软件有限公司 | Sound processing method, device, electronic equipment and storage medium |
CN113707167A (en) * | 2021-08-31 | 2021-11-26 | 北京地平线信息技术有限公司 | Training method and training device for residual echo suppression model |
US11984110B2 (en) * | 2022-03-07 | 2024-05-14 | Mediatek Singapore Pte. Ltd. | Heterogeneous computing for hybrid acoustic echo cancellation |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1809105B (en) * | 2006-01-13 | 2010-05-12 | 北京中星微电子有限公司 | Dual-microphone speech enhancement method and system applicable to mini-type mobile communication devices |
US8175871B2 (en) * | 2007-09-28 | 2012-05-08 | Qualcomm Incorporated | Apparatus and method of noise and echo reduction in multiple microphone audio systems |
US8811601B2 (en) * | 2011-04-04 | 2014-08-19 | Qualcomm Incorporated | Integrated echo cancellation and noise suppression |
US9768829B2 (en) * | 2012-05-11 | 2017-09-19 | Intel Deutschland Gmbh | Methods for processing audio signals and circuit arrangements therefor |
CN104519212B (en) * | 2013-09-27 | 2017-06-20 | 华为技术有限公司 | A kind of method and device for eliminating echo |
-
2016
- 2016-06-07 CN CN201610399409.4A patent/CN107483761B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN107483761A (en) | 2017-12-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107483761B (en) | A kind of echo suppressing method and device | |
CN105144674B (en) | Multi-channel echo is eliminated and noise suppressed | |
CN101562669B (en) | Method of adaptive full duplex full frequency band echo cancellation | |
CN101179295B (en) | Echo canceller and communication audio processing apparatus | |
KR101469739B1 (en) | A device for and a method of processing audio signals | |
CN105825864B (en) | Both-end based on zero-crossing rate index is spoken detection and echo cancel method | |
CN111768796B (en) | Acoustic echo cancellation and dereverberation method and device | |
EP1287672B1 (en) | Method and device for acoustic echo cancellation combined with adaptive beamforming | |
CN108376548B (en) | Echo cancellation method and system based on microphone array | |
CN106454642B (en) | Adaptive sub-band audio feedback suppression methods | |
CN108353107A (en) | The double talk detection eliminated for acoustic echo | |
CN110169041A (en) | Sound Echo cancellation | |
US9947337B1 (en) | Echo cancellation system and method with reduced residual echo | |
US20090316923A1 (en) | Multichannel acoustic echo reduction | |
EP3282678B1 (en) | Signal processor with side-tone noise reduction for a headset | |
WO2021110175A1 (en) | Echo cancellation method and device | |
JPH10190848A (en) | Method and system for canceling acoustic echo | |
US9313573B2 (en) | Method and device for microphone selection | |
US9191519B2 (en) | Echo suppressor using past echo path characteristics for updating | |
JP2003234679A (en) | Gain control method for executing acoustic echo cancellation and suppression | |
WO2015086229A9 (en) | Echo cancellation | |
CN111742541B (en) | Acoustic echo cancellation method, acoustic echo cancellation device and storage medium | |
JP2004537219A (en) | Echo canceller with nonlinear echo suppressor for harmonic calculation | |
CN111145771A (en) | Voice signal processing method, processing device, terminal and storage medium thereof | |
CN101292508B (en) | Acoustic echo canceller |
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 |