CN106161820B - A kind of interchannel decorrelation method for stereo acoustic echo canceler - Google Patents
A kind of interchannel decorrelation method for stereo acoustic echo canceler Download PDFInfo
- Publication number
- CN106161820B CN106161820B CN201510180915.XA CN201510180915A CN106161820B CN 106161820 B CN106161820 B CN 106161820B CN 201510180915 A CN201510180915 A CN 201510180915A CN 106161820 B CN106161820 B CN 106161820B
- Authority
- CN
- China
- Prior art keywords
- frequency
- value
- signal
- obtains
- amplitude
- 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
Landscapes
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
The present invention provides a kind of interchannel decorrelation methods for stereo acoustic echo canceler, comprising: step 1) distal end microphone picks up acoustic reference signal x1(n) and x2(n)；Frequency spectrum X is obtained by Fourier transformation1(k) and X2(k)；Step 2) is to frequency spectrum X1(k) and X2(k) high frequency decorrelative transformation is carried out；The time domain reference signal x that obtains that treated1d(n) and x2d(n)；Step 3) is to time domain reference signal x1(n) low frequency decorrelative transformation is carried out, three sinusoidal signals are constructed；Above three sinusoidal signal is added to x by time domain addition by step 4)1d(n) in；Obtain x1′(n)；x2' (n) is x2d(n)；The signal x that step 5) obtains step 4)1' (n) and x2' (n) carries out adaptive channel identification by adaptive algorithm, realizes echo cancellor.The Inter-channel Correlation of boombox signal is effectively reduced in method of the invention, improves the imbalance performance of stereo echo canceller system.
Description
Technical field
It is the present invention relates to acoustic echo technology for eliminating field, in particular to a kind of for the logical of stereo acoustic echo canceler
Decorrelation method between road.
Background technique
In hands-free communication system, due to acoustical signal closed loop transmission and there are acoustic echos.Such issues that in mobile phone
It is generally existing in the scenes such as communication, TeleConference Bridge, video conferencing system.Acoustic echo can reduce the communication quality of system,
Therefore it needs to be eliminated, generally uses adaptive channel identification algorithm (bibliography [1]: Benesty J, Kellermann
K et al.A perspective on stereophonic echo cancellation.Berlin,Heidelberg:
Springer-Verlag Berlin Heidelberg, 2011:10.) realize echo elimination.In stereo communication system,
There are strong Inter-channel Correlations for boombox signal in near-end room, this leads to the receipts of adaptive channel identification algorithm
Hold back that value is not unique, so that the filter for making stereo acoustic echo eliminate system is lacked of proper care larger (bibliography [2]: Benesty
J,Morgan D R et al.A better understanding and an improved solution to the
specific problems of stereophonic acoustic echo cancellation.IEEE
Trans.Speech and Audio Processing,1998,6(2):156-165.)。
In order to improve imbalance performance, need to carry out decorrelation pretreatment to two channel Near end speaker signals in advance；Meanwhile
In order to guarantee communication quality, respective handling should not introduce apparent distortion.Have both at home and abroad at present it is some can be used for it is three-dimensional several
Learn the interchannel decorrelation method of echo cancellor；Principle is a part of other signals to be added in original signal or directly to original
Beginning signal is converted；Wherein most representative is classical non-linear transform algorithm (bibliography [2]), and this method is original
The nonlinear transformation ingredient of a part of original signal is added in signal.The defect of these conventional methods is that quality distortion is big, and
Decorrelation effect is stronger, causes quality distortion bigger.In order to obtain better compromise between decorrelation effect and quality distortion,
In recent years, domestic and international researcher starts with psychologic acoustics effect to carry out interchannel decorrelative transformation, to obtain smaller sound
Matter distortion；Auditory masking effect (bibliography [3]: Valin J M.Perceptually-motivated in psychologic acoustics
nonlinear channel decorrelation for stereo acoustic echo cancellation,In
Proc.Joint Workshop Hands-Free Speech Communication and Microphone Arrays,
Trento, May 2008:188-191.) and fundamental frequency loss effect (bibliography [4]: Cecchi S, Romoli L et al.A
combined psychoacoustic approach for stereo acoustic echo cancellation.IEEE
Trans.Audio, Speech and Language Processing, 2011,19 (6): 1530-1539.) for carrying out low frequency
Decorrelative transformation.
Compared with the adaptive channel discrimination method of early stage, the method based on psychologic acoustics effect can be smaller in quality distortion
While obtain better decorrelation effect.But for eliminating the more demanding occasion of echo, current is imitated based on psychologic acoustics
The method answered is not able to satisfy higher quality distortion requirement, and the calculating process of this method is complex.
Summary of the invention
It is an object of the invention to overcome currently based on the method for psychologic acoustics effect above-mentioned present on elimination echo
Defect proposes a kind of interchannel decorrelation method for stereo acoustic echo canceler, and this method can be effectively reduced vertical
The Inter-channel Correlation of body sound loudspeaker signal improves the imbalance performance of stereo echo canceller system, it ensure that lesser
Quality distortion, and realize that simply computation complexity is low.
In order to achieve the above-mentioned object of the invention, the present invention proposes that a kind of interchannel for stereo acoustic echo canceler goes phase
Pass method, which comprises
Step 1) distal end microphone pick-up of acoustic signals obtains stereo reference signal x1(n) and x2(n)；Simultaneously to x1(n)
And x2(n) Fourier transformation is carried out, frequency spectrum X is obtained1(k) and X2(k)；
Step 2) is to frequency spectrum X1(k) and X2(k) high frequency decorrelative transformation is carried out；The time domain reference signal x that obtains that treated1d
(n) and x2d(n)；
Step 3) is to three-dimensional acoustic reference signal x1(n) low frequency decorrelative transformation is carried out, three sinusoidal signals: s are constructed1
(n)、s2(n) and s3(n)；
Step 4) passes through time domain addition for three sinusoidal signal s1(n)、s2(n) and s3(n) it is added to x1d(n) in；?
To x '1(n)；x′2It (n) is x2d(n)；The x '1(n) and x '2It (n) is the stereo signal after Whole frequency band decorrelative transformation；
The signal x ' that step 5) obtains step 4)1(n) and x '2(n) adaptive channel is carried out by adaptive algorithm to distinguish
Know, realizes echo cancellor.
In above-mentioned technical proposal, the step 2) specifically:
Simultaneously to frequency spectrum X1(k) and X2(k) high frequency decorrelative transformation is carried out:
Wherein, β (k) is nonlinear weight coefficient, it is contemplated that the relationship of the phase sensitive following frequency of human auditory system,
β (k) value are as follows:
Wherein, f (k) is linear frequency value corresponding to discrete frequency indices k,NFFT is in Fu
The length of leaf transformation；fsFor the sample frequency of signal；β (k) is in fsIt is maximized at/2；
Judge that analog frequency f (k) corresponding to k is greater than f0It is whether true, the f0It is high frequency or low frequency for division signals
Critical value；If a determination be made that negative, 0 is set by Θ value, otherwise, sets Θ value to the value of sinusoidal variations；
To frequency spectrum X1d(k) and X2d(k) inverse Fourier transform is carried out, the time domain reference after high frequency decorrelative transformation is obtained
Signal x1d(n) and x2d(n)。
In above-mentioned technical proposal, the f0Value are as follows: f0=1kHz.
In above-mentioned technical proposal, the step 3) is specifically included:
Step 3-1) calculate reference signal x1(n) weighted auto-correlation function ψw(τ)；
x1(n) average magnitude difference function φ (τ) is defined as:
φ (τ)=E | x1(n)-x1(n+τ)|} (7)
Wherein, τ is time delay, and desired operation is sought in E { } expression；
x1(n) auto-correlation function is calculated as follows:
ψ (τ)=E { x1(n)x1(n+τ)}. (8)
Weighted auto-correlation function are as follows:
ψw(τ)=ψ (τ)/[φ (τ)+ε], (9)
Wherein ε is the fixed value for preventing quotient from dissipating and introducing, ε=1；
Step 3-2) calculate fundamental frequency cycles estimated value τ0；
τ0={ τ: ψw(τ)=max (ψw(τ))；fs/350≤τ≤fs/60} (10)
Wherein, max () expression is maximized；
Step 3-3) recalculate fundamental frequency estimated value f0；
To with maximum value ψw(τ0) centered on adjacent 3 ψw(τ) value carries out interpolation, then searches again for maximum value and obtains essence
Spend higher fundamental frequency cycles estimated value T0, then fundamental frequency value are as follows:
f0=1/T0 (11)
Step 3-4) retrieval reference signal frequency spectrum X1(k), it carries out amplitude weighting and obtains the amplitude A of three low order harmonics ingredientsi
And first phase
k0Indicate frequency indices corresponding to fundamental frequency, then two, harmonics corresponds respectively to 2k three times0And 3k0；Utilize three frequencies
Rate index value retrieves X1(k), the amplitude and first phase of fundamental frequency and two, harmonics three times are respectively obtained, amplitude weighting is carried out and obtains:
Wherein, A1、And σ1For fundamental frequency amplitude, first phase and the amplitude weighting factor, A2、And σ2For secondary harmonics amplitude,
First phase and the amplitude weighting factor, A3、And σ3For for the amplitude of harmonics three times, first phase and the amplitude weighting factor；Angle { } table
Show and phase function, σ are taken to plural numberi(i=1,2,3) is in [0,1] interior value, it is preferred that σ1=0.1, σ2=0.2, σ3=0.3；
Step 3-5) three sinusoidal signal s of construction1(n)、s2(n) and s3(n):
Wherein, fi=if0, i=1,2,3.
In above-mentioned technical proposal, the adaptive algorithm are as follows: least fibre method, normalized least fibre method or frequency domain are most
Small equal method.
The present invention has the advantages that
1, method of the invention utilizes the nonlinear Phase Modulation of time-varying to traditional nonlinear transformation decorrelation in high band
Method improves, thus obtains better high frequency decorrelation effect；
2, method of the invention be based on psychologic acoustics compose advantageous effect, near three low order harmonics ingredients Implantation Energy compared with
Weak sinusoidal signal is to carry out low frequency decorrelative transformation, it ensure that lesser quality distortion；
3, method of the invention is injected sinusoidal signal in original signal using time domain addition, and algorithm is made to have lower fortune
Calculation amount；
4, method of the invention is lost by adjusting the energy size of injected sinusoidal signal in decorrelation effect and sound quality
Different degrees of compromise is carried out between true, therefore can be according to the requirement of concrete application scene by sinusoidal signal in practical application
Energy is arranged to different value, such as those occasions of less demanding to sound quality, can increase injected sinusoidal signal
Energy is to obtain better decorrelation effect.
Detailed description of the invention
Fig. 1 is that stereo acoustic echo eliminates model general illustration；
Fig. 2 is flow diagram of the present invention for the interchannel decorrelation method of stereo acoustic echo canceler；
Fig. 3 is that the obtained interchannel in example 1 using method and existing non-linear transform algorithm of the invention is relevant
Index contrast figure；
Fig. 4 is in example 1 using the obtained imbalance performance of method and existing non-linear transform algorithm of the invention
Comparison diagram；
Fig. 5 is the obtained voice quality sense using method Yu existing non-linear transform algorithm of the invention in example 1
Know the comparison diagram of assessment.
Specific embodiment
As shown in Figure 1, in hands-free communication system, the acoustical signal of distal end speaker by distal end stereo microphone
It receives, obtains the stereo reference signal x in two channels1(n) and x2(n).Stereo reference signal x1(n) and x2(n) through past phase
Pretreatment is closed, the signal x that obtains that treated1' (n) and x2'(n).Then pass through line transmission to proximal end, x1' (n) and x2'(n)
It is played again by the stereophonic loud-speaker system of proximal end.The microphone channel signal of proximal end two is respectively d1(n) and d2(n), i.e. proximal end
The signal that microphone receives, is divided into two parts: a part is x1' (n) and x2' after (n) played back by Near end speaker,
Acoustical signal is propagated in proximal void, after reflection and scattering in near-end room, reaches the signal of proximal end microphone；Separately
A part is ambient noise and voice in the scene of proximal end, i.e. near end signal.Proximal end microphone signal d1(n) and d2(n) in
First part's signal can return to distal end by line transmission again, and be reset in the scene of distal end by far-end loudspeaker.So,
Speaker in the scene of distal end has just heard the sound of a part oneself, and here it is the acoustic echos that needs are eliminated.
If signal vector corresponding to two channel speakers signals is respectively as follows:
x1' (n)=[x1′(n-L+1) x1′(n-L+2) ... x1′(n)]
x2' (n)=[x2′(n-L+1) x2′(n-L+2) ... x2′(n)]
The acoustic echo showed in the scene of distal end mainly due to caused by the acoustic propagation in the scene of proximal end,
It is recognized in the scene of proximal end using adaptive channel and carries out acoustic echo elimination.h12(n) and h22(n) proximal end scene is respectively indicated
In two loudspeaker channels to left microphone room impulse response,WithIt is then adaptive channel identification respectively
Algorithm is to h12(n) and h22(n) estimation；h21(n) and h11(n) two loudspeaker channels are respectively indicated in the scene of proximal end to right biography
The room impulse response of sound device,WithIt is then adaptive channel identification algorithm respectively to h21(n) and h11(n) estimate
Meter.
Adaptive algorithm passes through
Obtain the residue signal e after echo cancellor1(n) and e2(n)；e1(n) and e2(n) by vertical in the scene of distal end
Body sound loudspeaker plays, and can so eliminate the acoustic echo occurred in the scene of distal end.
Filter imbalance is defined as:
Wherein, | | | |2Indicate 2 norms.
If pre-processing without decorrelation, biggish filter imbalance value η (n) will be present.That is, adaptive channel is distinguished
Knowing algorithm, there are biggish filter evaluated errors.Especially when physical pulse response changes, big filter is lost
Tune will lead to adaptive tracing and slow or even can not accurately track its variation, to make echo cancellor effect severe exacerbation.
The reason is that if being pre-processed without decorrelation,
Due to stereo reference signal x1(n) and x2(n) the same speaker in the scene of distal end, between them
With very strong Inter-channel Correlation.This strong correlation will make the convergency value of adaptive channel identification algorithm not unique, real
The sef-adapting filter that border converges to when applyingIt may and be not equal to our desired physics
Transmission function hij(n), i=1,2；J=1,2, it lacks of proper care so as to cause big filter.
In order to solve this problem, improve imbalance performance, usually in reference signal x1(n) and x2(n) it is transmitted to proximal end scene
Decorrelation pretreatment is carried out to it before, so that filter imbalance be effectively reduced, improves stereo acoustic echo and eliminates performance.
The present invention will be further described in detail in the following with reference to the drawings and specific embodiments.
As shown in Fig. 2, a kind of interchannel decorrelation method in the elimination of stereo acoustic echo, specific steps are such as
Under:
Step 1) distal end microphone pick-up of acoustic signals obtains stereo reference signal x1(n) and x2(n)；Simultaneously to x1(n)
And x2(n) Fourier transformation (FFT) is carried out, obtains frequency spectrum X1(k) and X2(k)；
Step 2) is to frequency spectrum X1(k) and X2(k) high frequency decorrelative transformation is carried out；The time domain reference signal x that obtains that treated1d
(n) and x2d(n)；It specifically includes:
Simultaneously to frequency spectrum X1(k) and X2(k) high frequency decorrelative transformation is carried out:
Wherein, β (k) is nonlinear weight coefficient, it is contemplated that the relationship of the phase sensitive following frequency of human auditory system,
β (k) value are as follows:
Wherein, f (k) is linear frequency value corresponding to discrete frequency indices k,NFFT is in Fu
The length of leaf transformation；fsFor the sample frequency of signal；β (k) is in fsIt is maximized at/2；
Judge that analog frequency f (k) corresponding to k is greater than f0It is whether true, the f0It is high frequency or low frequency for division signals
Critical value；Preferably, f0=1kHz；If a determination be made that negative, 0 is set by Θ value, otherwise, is set Θ value to
The value of sinusoidal variations；
To frequency spectrum X1d(k) and X2d(k) carry out inverse Fourier transform (IFFT), obtain after high frequency decorrelative transformation when
Domain reference signal x1d(n) and x2d(n)。
Step 3) is to three-dimensional acoustic reference signal x1(n) low frequency decorrelative transformation is carried out, three sinusoidal signals are constructed；Specifically
Include:
Step 3-1) calculate stereo reference signal x1(n) weighted auto-correlation function ψw(τ)；
x1(n) average magnitude difference function (AMDF) φ (τ) is defined as:
φ (τ)=E | x1(n)-x1(n+τ)|} (7)
Wherein, τ is time delay, and desired operation is sought in E { } expression；
x1(n) auto-correlation function is calculated as follows:
ψ (τ)=E { x1(n)x1(n+τ)}. (8)
Weighted auto-correlation function is are as follows:
ψw(τ)=ψ (τ)/[φ (τ)+ε], (9)
Wherein ε is the fixed value for preventing quotient from dissipating and introducing, preferred value ε=1.
Step 3-2) calculate fundamental frequency cycles estimated value τ0；
Weighted auto-correlation function ψwThe maximum value ψ of (τ)w,maxCorresponding time delay number τ0The as estimated value of fundamental frequency cycles；
The value range of fundamental frequency is limited to [60,350] Hz, this, which is equivalent to, limits τ0Search range, it may be assumed that
τ0={ τ: ψw(τ)=max (ψw(τ))；fs/350≤τ≤fs/60} (10)
Wherein, max () expression is maximized.
Step 3-3) recalculate fundamental frequency estimated value f0；
In order to improve the detection accuracy of fundamental frequency, further to maximum value ψw(τ0) centered on adjacent 3 ψw(τ) value into
Then row interpolation searches again for maximum value and obtains the higher fundamental frequency cycles estimated value T of precision0, then fundamental frequency value are as follows:
f0=1/T0 (11)
Step 3-4) retrieval reference signal frequency spectrum X1(k), it carries out amplitude weighting and obtains the amplitude A of three low order harmonics ingredientsi
And first phase
k0Indicate frequency indices corresponding to fundamental frequency, then two, harmonics corresponds respectively to 2k three times0And 3k0；Utilize three frequencies
Rate index value retrieves X1(k), the amplitude and first phase of fundamental frequency and two, harmonics three times are respectively obtained, amplitude weighting is carried out and obtains:
Wherein, A1、And σ1For fundamental frequency amplitude, first phase and the amplitude weighting factor, A2、And σ2For secondary harmonics amplitude,
First phase and the amplitude weighting factor, A3、And σ3For for the amplitude of harmonics three times, first phase and the amplitude weighting factor；Angle { } table
Show and phase function, σ are taken to plural numberi(i=1,2,3) is in [0,1] interior value, it is preferred that σ1=0.1, σ2=0.2, σ3=0.3.
Step 3-5) three sinusoidal signal s of construction1(n)、s2(n) and s3(n):
Wherein, fi=if0, i=1,2,3.
Step 4) passes through time domain addition for three sinusoidal signal s1(n)、s2(n) and s3(n) it is added to x1d(n) in；?
To x '1(n)；x′2It (n) is x2d(n)；x′1(n) and x '2(n) it is the stereo signal after Whole frequency band decorrelative transformation:
The signal x ' that step 5) obtains step 4)1(n) and x '2(n) it is carried out by normalized least fibre method adaptive
Channel identification, realizes echo cancellor；
Within n-th of sampling period, the sef-adapting filter in n-th of sampling period is utilizedTo x '1(n) and x '2
(n) it filters, obtains the echo signal y at microphonej(n), it may be assumed that
Again in microphone signal dj(n) echo signal is eliminated in, obtains residue signal ej(n):
ej(n)=dj(n)-yj(n) (16)
Then sef-adapting filter is updated, obtains the sef-adapting filter in (n+1)th sampling period
Wherein, xi' (n)=[xi′(n-L+1) xi′(n-L+2) ... xi′(n)]。
For next sampling period, then continue to carry out adaptive tracing from formula (17) to formula (19), until signal terminates.
The normalized least fibre method (NLMS) can be other adaptive algorithms: least fibre method (LMS) or frequency
Domain least fibre method (FBLMS).
Below with reference to example, the present invention will be further described.
Example 1: near-end room is having a size of 4 × 3 × 3m3, two Near end speakers be located at { 1,2,1.2 } m and 3,2,
1.2}m；The proximal end microphone being located at { 1.8,1,1.2 } m is investigated, the acoustical signal that it is picked up regards proximal end microphone letter as
Number d (n).Signal-to-noise ratio is used to simulate near end signal for the white Gaussian noise of 30dB.Sample rate takes 16kHz, using NLMS algorithm into
The identification of row adaptive channel.Method of the invention and traditional nonlinear transformation decorrelation method (NLT) are subjected to performance comparison.
Interchannel coherence factor (the Inter-Channel Coherence of two channel signal x1 (n) and x2 (n)
Coefficient, ICCC) is defined as:
ICCC is in [0,1] interior value, and value is smaller, and inter-channel coherence is weaker, and decorrelation effect is better.As shown in figure 4,
In the processing result of three methods, the ICCC value of treated by means of the present invention two signals is minimum.
In learning curve of lacking of proper care, it is shorter to reach stable state time-consuming, indicates that convergence rate is faster；Its value is smaller, and identification precision is got over
It is high.If Fig. 4 shows, in the processing result of three methods, treated by means of the present invention misregistration signal best performance.
Perceptual evaluation of speech quality (Perceptual Evaluation of Speech Quality, PESQ) is to a variety of
Human hearing characteristic modeling, by subjective sound quality objective quantification.PESQ value between -0.5 to 4.5, value is bigger, indicates distortion
It is smaller.As shown in figure 5, in the processing result of three methods, the PESQ value ratio of treated by means of the present invention signal
The PESQ value of signal is big after NLT processing.
Claims (5)
1. a kind of interchannel decorrelation method for stereo acoustic echo canceler, which comprises
Step 1) distal end microphone pick-up of acoustic signals obtains stereo reference signal x1(n) and x2(n)；Simultaneously to x1(n) and x2
(n) Fourier transformation is carried out, frequency spectrum X is obtained1(k) and X2(k)；
Step 2) is to frequency spectrum X1(k) and X2(k) high frequency decorrelative transformation is carried out；The time domain reference signal x that obtains that treated1d(n) and
x2d(n)；
Step 3) is to three-dimensional acoustic reference signal x1(n) low frequency decorrelative transformation is carried out, three sinusoidal signals: s are constructed1(n)、s2
(n) and s3(n)；
Step 4) passes through time domain addition for three sinusoidal signal s1(n)、s2(n) and s3(n) it is added to x1d(n) in；It obtains
x′1(n)；x′2It (n) is x2d(n)；x′1(n) and x '2It (n) is the stereo signal after Whole frequency band decorrelative transformation；
The signal x ' that step 5) obtains step 4)1(n) and x '2(n) adaptive channel identification is carried out by adaptive algorithm, it is real
Existing echo cancellor.
2. the interchannel decorrelation method according to claim 1 for stereo acoustic echo canceler, which is characterized in that
The step 2) specifically:
Simultaneously to frequency spectrum X1(k) and X2(k) high frequency decorrelative transformation is carried out:
Wherein, β (k) is nonlinear weight coefficient, it is contemplated that the relationship of the phase sensitive following frequency of human auditory system, β (k)
Value are as follows:
Wherein, f (k) is linear frequency value corresponding to discrete frequency indices k,NFFT is Fourier transformation
Length；fsFor the sample frequency of signal；β (k) is in fsIt is maximized at/2；
Judge that analog frequency f (k) corresponding to k is greater than f0It is whether true, the f0It is facing for high frequency or low frequency for division signals
Dividing value；If a determination be made that negative, 0 is set by Θ value, otherwise, sets Θ value to the value of sinusoidal variations；
To frequency spectrum X1d(k) and X2d(k) inverse Fourier transform is carried out, the time domain reference signal after high frequency decorrelative transformation is obtained
x1d(n) and x2d(n)。
3. the interchannel decorrelation method according to claim 2 for stereo acoustic echo canceler, which is characterized in that
The f0Value are as follows: f0=1kHz.
4. the interchannel decorrelation method according to claim 2 for stereo acoustic echo canceler, which is characterized in that
The step 3) specifically includes:
Step 3-1) calculate reference signal x1(n) weighted auto-correlation function ψw(τ)；
x1(n) average magnitude difference function φ (τ) is defined as:
φ (τ)=E | x1(n)-x1(n+τ)|} (7)
Wherein, τ is time delay, and desired operation is sought in E { } expression；
x1(n) auto-correlation function is calculated as follows:
ψ (τ)=E { x1(n)x1(n+τ)}. (8)
Weighted auto-correlation function are as follows:
ψw(τ)=ψ (τ)/[φ (τ)+ε], (9)
Wherein ε is the fixed value for preventing quotient from dissipating and introducing, ε=1；
Step 3-2) calculate fundamental frequency cycles estimated value τ0；
τ0={ τ: ψw(τ)=max (ψw(τ))；fs/350≤τ≤fs/60} (10)
Wherein, max () expression is maximized；
Step 3-3) recalculate fundamental frequency estimated value f0；
To with maximum value ψw(τ0) centered on adjacent 3 ψw(τ) value carries out interpolation, then searches again for maximum value and obtains precision more
High fundamental frequency cycles estimated value T0, then fundamental frequency value are as follows:
f0=1/T0 (11)
Step 3-4) retrieval reference signal frequency spectrum X1(k), it carries out amplitude weighting and obtains the amplitude A of three low order harmonics ingredientsiWith it is first
Phase
k0Indicate frequency indices corresponding to fundamental frequency, then two, harmonics corresponds respectively to 2k three times0And 3k0, utilize three frequency ropes
Draw value retrieval X1(k), the amplitude and first phase of fundamental frequency and two, harmonics three times are respectively obtained, amplitude weighting is carried out and obtains:
Wherein, A1、And σ1For fundamental frequency amplitude, first phase and the amplitude weighting factor, A2、And σ2Amplitude, first phase for secondary harmonics
With the amplitude weighting factor, A3、And σ3For for the amplitude of harmonics three times, first phase and the amplitude weighting factor；Angle { } expression pair
Plural number takes phase function, σi, i=1,2,3 in [0,1] interior value, σ1=0.1, σ2=0.2, σ3=0.3；
Step 3-5) three sinusoidal signal s of construction1(n)、s2(n) and s3(n):
Wherein, fi=if0, i=1,2,3.
5. the interchannel decorrelation method according to claim 4 for stereo acoustic echo canceler, which is characterized in that
The adaptive algorithm are as follows: least fibre method, normalized least fibre method or frequency domain least fibre method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510180915.XA CN106161820B (en) | 2015-04-16 | 2015-04-16 | A kind of interchannel decorrelation method for stereo acoustic echo canceler |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510180915.XA CN106161820B (en) | 2015-04-16 | 2015-04-16 | A kind of interchannel decorrelation method for stereo acoustic echo canceler |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106161820A CN106161820A (en) | 2016-11-23 |
| CN106161820B true CN106161820B (en) | 2019-04-23 |
Family
ID=58058019
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510180915.XA Active CN106161820B (en) | 2015-04-16 | 2015-04-16 | A kind of interchannel decorrelation method for stereo acoustic echo canceler |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106161820B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107452398B (en) * | 2017-08-09 | 2021-03-16 | 深圳创维数字技术有限公司 | Echo acquisition method, electronic device and computer readable storage medium |
| CN109509482B (en) * | 2018-12-12 | 2022-03-25 | 北京达佳互联信息技术有限公司 | Echo cancellation method, echo cancellation device, electronic apparatus, and readable medium |
| CN111756943B (en) * | 2020-06-05 | 2021-04-20 | 中原工学院 | Self-adaptive stereo acoustic echo cancellation method for sub-band with increased complex value and proportional value |
| CN111726464B (en) * | 2020-06-29 | 2021-04-20 | 珠海全志科技股份有限公司 | Multichannel echo filtering method, filtering device and readable storage medium |
| CN111755020B (en) * | 2020-08-07 | 2023-02-28 | 南京时保联信息科技有限公司 | Stereo echo cancellation method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101010723A (en) * | 2004-08-25 | 2007-08-01 | 杜比实验室特许公司 | Multichannel decorrelation in spatial audio coding |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9357307B2 (en) * | 2011-02-10 | 2016-05-31 | Dolby Laboratories Licensing Corporation | Multi-channel wind noise suppression system and method |
-
2015
- 2015-04-16 CN CN201510180915.XA patent/CN106161820B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101010723A (en) * | 2004-08-25 | 2007-08-01 | 杜比实验室特许公司 | Multichannel decorrelation in spatial audio coding |
Non-Patent Citations (1)
| Title |
|---|
| "一种基于谱优势的混合立体声声学回声抵消方法";杨鹤飞等：;《声学技术》;20131231;第32卷(第6期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106161820A (en) | 2016-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sridhar et al. | ICASSP 2021 acoustic echo cancellation challenge: Datasets, testing framework, and results | |
| CN107123430B (en) | Echo cancellation method, device, conference tablet and computer storage medium | |
| EP3791565B1 (en) | Method and apparatus utilizing residual echo estimate information to derive secondary echo reduction parameters | |
| CN103428385B (en) | For handling the method for audio signal and circuit arrangement for handling audio signal | |
| CN106161820B (en) | A kind of interchannel decorrelation method for stereo acoustic echo canceler | |
| Cutler et al. | INTERSPEECH 2021 Acoustic Echo Cancellation Challenge. | |
| Carbajal et al. | Multiple-input neural network-based residual echo suppression | |
| Zhang et al. | FT-LSTM based complex network for joint acoustic echo cancellation and speech enhancement | |
| CN108447496B (en) | Speech enhancement method and device based on microphone array | |
| US11380312B1 (en) | Residual echo suppression for keyword detection | |
| Habets et al. | Joint dereverberation and residual echo suppression of speech signals in noisy environments | |
| Schwartz et al. | Joint estimation of late reverberant and speech power spectral densities in noisy environments using Frobenius norm | |
| Ivry et al. | Nonlinear acoustic echo cancellation with deep learning | |
| CN110211602A (en) | Intelligent sound enhances communication means and device | |
| CN113241085A (en) | Echo cancellation method, device, equipment and readable storage medium | |
| Koldovský et al. | Noise reduction in dual-microphone mobile phones using a bank of pre-measured target-cancellation filters | |
| CN113409810B (en) | Echo cancellation method for joint dereverberation | |
| CN112489680B (en) | Evaluation method and device of acoustic echo cancellation algorithm and terminal equipment | |
| CN112133324A (en) | Call state detection method, device, computer system and medium | |
| CN115620737A (en) | Voice signal processing device, method, electronic equipment and sound amplification system | |
| Garre et al. | An Acoustic Echo Cancellation System based on Adaptive Algorithm | |
| Zhang et al. | Deep learning for joint acoustic echo and acoustic howling suppression in hybrid meetings | |
| CN110246516A (en) | The processing method of small space echo signal in a kind of voice communication | |
| Sun et al. | Time-frequency Complex Mask Network for Echo Cancellation and Noise Suppression | |
| US20230344941A1 (en) | Method for echo cancellation, echo cancellation device and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |