CN107105111A - One kind combination proportional affine projection echo cancel method of step-length - Google Patents
One kind combination proportional affine projection echo cancel method of step-length Download PDFInfo
- Publication number
- CN107105111A CN107105111A CN201710155109.6A CN201710155109A CN107105111A CN 107105111 A CN107105111 A CN 107105111A CN 201710155109 A CN201710155109 A CN 201710155109A CN 107105111 A CN107105111 A CN 107105111A
- Authority
- CN
- China
- Prior art keywords
- sef
- current time
- adapting filter
- mrow
- signal
- 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.)
- Granted
Links
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 title claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims description 28
- 230000003044 adaptive effect Effects 0.000 claims description 17
- 238000005070 sampling Methods 0.000 claims description 6
- 230000017105 transposition Effects 0.000 claims description 6
- 230000005611 electricity Effects 0.000 claims description 2
- 238000002474 experimental method Methods 0.000 description 9
- 230000004044 response Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000010606 normalization Methods 0.000 description 5
- 230000006854 communication Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
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
- H04M9/085—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 digital techniques
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L2021/02082—Noise filtering the noise being echo, reverberation of the speech
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Filters That Use Time-Delay Elements (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
One kind combination proportional affine projection echo cancel method of step-length, its step is as follows：Collection desired signal d (n) at A, the voice signal x (n) that collection distal end is transmitted at the telelecture and telephone microphone；B, using voice signal x (n) as the input of sef-adapting filter, it is output as the estimate of echo signalC, echo cancelltion-subtract d (n)Obtained error signal e (n) sends distal end to as the clean signal eliminated after echo；D, the error vector for calculating sef-adapting filter；E, the right value update increment Delta w (n) for calculating sef-adapting filter；F, calculating combination step size mu (n)；G, weight vector w (n+1)=w (n)+μ (n) the Δ w (n) for updating sef-adapting filter；H, make n=n+1, the step of repeating A, B, C, D, E, F, G, until end of conversation.The present invention has the advantages that fast convergence rate, while steady-state error is low.
Description
Technical field
The invention belongs to the adaptive echo technology for eliminating field of telephone communication.
Background technology
In telephone communication service, every needs use the occasion of microphone and loudspeaker simultaneously, all inevitably
There is acoustic echo, it is to influence the main factor of speech quality.In communication process, the sound of other side speaker passes through electricity
The loudspeaker of words is played back, direct (without any reflection) or indirectly (once or many by the object in house or house
Secondary reflection) picked up by the microphone (microphone) of phone, then distant place speaker is returned to, it is heard word before oneself, i.e.,
Acoustic echo.Propagation path of the sound from loudspeaker to microphone is called echo channel, its impulse response can with one arrange to
Amount is represented.
Acoustic echo influences speech quality, and various echo cancellation technologies are proposed to this researcher.At present, certainly
It is listener's echo technology for eliminating conventional in the world to adapt to echo cancellation technology.In adaptive echo elimination, it will be raised from phone
The voice signal played back is gathered at sound device as the input signal of sef-adapting filter, the output of sef-adapting filter is obtained
Signal is exactly the estimate of echo, and the signal for then picking up the microphone of phone is subtracted just to be returned to far after the estimate of echo
The speaker of side.Adaptive filter algorithm is exactly the impulse response that sef-adapting filter is constantly adjusted with the change according to environment,
The impulse response of echo channel, and then the more preferable estimated echo of output energy of sef-adapting filter can be picked out, to reach
The adaptive purpose for eliminating echo.
Acoustical echo channel has openness mostly, i.e., most of coefficient of its impulse response is all close or equal to zero.
Therefore, in Sparse System, traditional adaptive filter algorithm such as normalization minimum mean-square calculation (NLMS), affine projection algorithm
(APA) convergence rate, which occurs, to be decreased obviously.Therefore, proportional algorithm is effectively improved convergence rate.Proportional algorithm
It is to distribute a coefficient for each weights of sef-adapting filter, and in each iterative process, the coefficient and corresponding weights
Amplitude size be directly proportional, i.e., the bigger weights of amplitude obtain bigger coefficient, so as to accelerate global convergence speed.Typical case
Proportional algorithm have improved proportional normalization minimum mean-square calculation (IPNLMS) algorithm and it is improved it is proportional it is affine throw
Shadow algorithm (IPAPA) algorithm, wherein, when input signal is coherent signal such as voice signal, IPAPA algorithms are than IPNLMS algorithm
There is faster convergence rate.Compared to IPAPA algorithms, memory IPAPA (MIPAPA) algorithm " C.Paleologu,
S.Ciochina,J.Benesty.An efficient proportionate affine projection algorithm
for echo cancellation[J].IEEE Signal Processing Letters,2010,17(2):165-168.”
It is more suitable for acoustic echo elimination, because it has faster convergence rate and lower computation complexity.However, MIPAPA algorithms
There is also the compromise problem between Fast Convergent and low steady-state error, the selection of fixed step size is attributed to.Precisely, it is one big
Step-length causes convergence of algorithm speed fast, but the steady-state error of algorithm is big；In turn, to cause algorithm to have small for a small step-length
Steady-state error, but slow convergence rate.
The content of the invention
The proportional affine projection echo cancel method of step-length is combined it is an object of the invention to provide one kind, this method can be simultaneously
Fast convergence rate and low steady-state error are obtained, so that echo cancellor effect is good.
The technical scheme adopted by the invention for realizing the object of the invention is that one kind combines the proportional affine projection echo of step-length
Removing method, its step is as follows：
1st, a kind of combination proportional affine projection echo cancel method of step-length, its step is as follows：
A, signal acquisition
The signal sampling that the loudspeaker of phone is sent, obtains current time n voice signal centrifugal pump x (n)；Meanwhile,
The signal sampling that the microphone of phone is induced, obtains current time n desired signal centrifugal pump d (n)；
B, echo signal estimation
Value x (n) by voice signal centrifugal pump x (n) at n to the n-L+1 moment, x (n-1) ..., x (n-L+1) are constituted and worked as
Preceding moment n sef-adapting filter input vector x (n), x (n)=[x (n), x (n-1) ..., x (n-L+1)]T, wherein, L is represented
The length of sef-adapting filter, L=512 or 1024, subscript T represents transposition；
Using voice signal centrifugal pump x (n) as input signal input adaptive wave filter, the adaptive of current time n is obtained
Answer the output signal of wave filter Wherein, w (n) is the weights of current n moment sef-adapting filter
Vector, its length is equal to L, and initial value is null vector；
C, echo cancelltion
A is walked to obtained current time n desired signal centrifugal pump d (n), current time n sef-adapting filter is subtracted
Output signalObtain current time n error signal e (n)；And by current time n error signal e (n), as disappearing
Except the clean signal after echo sends distal end to,；
D, the error vector of sef-adapting filter are calculated
D1, the sef-adapting filter input vector x (n) with current time n to moment n-M+1, x (n-1) ..., x (n-M+
1) sef-adapting filter, is constituted in current time n input matrix X (n), X (n)=[x (n), x (n-1) ..., x (n-M+
1)], wherein, M be affine projection exponent number, its span be 4~10；
D2, the desired signal d (n) with current time n to moment n-M+1, d (n-1) ..., d (n-M+1), are constituted current
Moment n desired signal vector d (n), d (n)=[d (n), d (n-1) ..., d (n-M+1)]T；
D3, current time n desired signal vector d (n) subtracted current time n input matrix X (n) transposition with from
The weight vector w (n) of adaptive filter product, obtains the error vector e (n) of current time n sef-adapting filter, e (n)
=d (n)-XT(n)w(n)；
E, sef-adapting filter right value update
E1, scale factor calculating
I-th of scale factor g of current time n sef-adapting filteri(n), drawn by following formula：
Wherein, wi(n) it is i-th element of the sef-adapting filter in the weight vector w (n) at moment at current time n, | |
||1Represent to ask vectorial 1- norms, κ is proportionality control parameter κ ∈ [- 1,1], and ε is regularization parameter, and its value is
0.0001~0.001；
By all proportions factor g of current time n sef-adapting filter1(n),g2(n)...gi(n)...,gL(n), structure
Into sef-adapting filter current time n column vector G (n),
G (n)=[g1(n),...gi(n)...,gL(n)]T；
E2, proportional intermediary matrix calculating
By the scale factor column vector G (0) of initial time n=0 sef-adapting filter diagonal matrix diag { G (0) }
It is multiplied with the input matrix X (0) of initial time n=0 sef-adapting filter, obtains initial time n=0 proportional middle square
Battle array P (0), P (0)=diag { G (0) } X (0)；
According to sef-adapting filter in current time n scale factor column vector G (n) and the adaptive-filtering at current time
Device input vector x (n), the P (n) of the proportional intermediary matrix for the current time n for obtaining L rows M row is calculated by following formula,
P (n)=[G (n) ⊙ x (n), P-1(n)],
Wherein, ⊙ represents point multiplication operation, P-1(n) the M-1 row before the intermediary matrix P (n-1) of previous moment (n-1)
Vector is constituted；
E3, sef-adapting filter right value update
The right value update increment Delta w (n) of current time n sef-adapting filter, is drawn by following formula：
Δ w (n)=P (n) [δ × I+XT(n)P(n)]-1E (n),
Wherein, I represents M × M unit matrix, []-1Finding the inverse matrix is represented, δ is stability contorting parameter, its value
For 0.001~0.01；
F, the calculating for combining step-length
F1, the combination step-length auxiliary parameter a (n) at current time are drawn as follows：
A (n)=a (n-1)+μa(μ1-μ2)[λ(n-1)(1-λ(n-1))+θ]e(n)xT(n)Δw(n-1)
Wherein, μaIt is the study control parameter for combining step-length, value is 100~300；μ1It is big step-length, its value is 1；μ2It is
Small step-length, its span is 0 ＜ μ2≤0.1；θ is combination step-length steadiness parameter, and value is 1 × 10-5；Δ w (n-1) is previous
The right value update increment of moment (n-1) sef-adapting filter, its initial value Δ w (0) is 0；Auxiliary parameter a (n) initial value is a
(0)=4；λ (n-1) is the step-length hybrid parameter of previous moment (n-1), and its initial value λ (0) is 0.
F2, step-length hybrid parameter calculating
Current time n step-length hybrid parameter λ (n) is drawn by the following method：
IfThen λ (n)=0；
IfThen λ (n)=1；
IfThen
Wherein, C is a mixing control constant, and span is 2~4；
F3, the calculating for combining step-length
Current time n combination step size mu (n) is drawn by following formula：
μ (n)=λ (n) μ1+(1-λ(n))μ2
G, sef-adapting filter right value update
Calculate what is obtained according to the step E right value update increment Delta w (n) and step F for calculating obtained sef-adapting filter
Step size mu (n) is combined, the weight vector w (n+1), w (n+1)=w (n) of the sef-adapting filter for obtaining subsequent time (n+1) is updated
+μ(n)Δw(n)；
H, repetition
N=n+1 is made, the step of repeating A, B, C, D, E, F, G, until end of conversation.
Compared with prior art, the beneficial effects of the invention are as follows：
In the adaptive incipient stage, hybrid parameter λ (n) value is close to 1, so that passing through μ (n)=λ (n) μ1+(1-
λ(n))μ2Combination step size mu (n) is obtained close to big step size mu1So that fast convergence rate of the invention；As algorithm is gradually restrained,
Hybrid parameter λ (n) value will be gradually reduced, and pass through μ (n)=λ (n) μ1+(1-λ(n))μ2Combination step size mu (n) is obtained also gradually to subtract
It is small；When algorithm has converged to stable state, hybrid parameter λ (n) value is equal to 0, then passes through μ (n)=λ (n) μ1+(1-λ(n))μ2
It is equal to small step size mu to combination step size mu (n)2So that the present invention has small steady-state error.
In a word, the present invention can obtain quick convergence rate and small steady-state error, so that the effect of echo cancellor
More preferably.
The present invention is described in detail with reference to the accompanying drawings and detailed description
Brief description of the drawings
Fig. 1 is the impulse response of the echo channel of emulation experiment of the present invention；
Fig. 2 is the voice signal of the telelecture broadcast of emulation experiment of the present invention；
Fig. 3 is the echo signal of the telephone microphone pickup of emulation experiment of the present invention；
Fig. 4 is the normalization imbalance curve of emulation experiment of the present invention.
Embodiment
Embodiment
A kind of embodiment of the present invention is, 1, a kind of combination proportional affine projection echo cancel method of step-length,
Its step is as follows：
A, signal acquisition
The signal sampling that the loudspeaker of phone is sent, obtains current time n voice signal centrifugal pump x (n)；Meanwhile,
The signal sampling that the microphone of phone is induced, obtains current time n desired signal centrifugal pump d (n)；
B, echo signal estimation
Value x (n) by voice signal centrifugal pump x (n) at n to the n-L+1 moment, x (n-1) ..., x (n-L+1) are constituted and worked as
Preceding moment n sef-adapting filter input vector x (n), x (n)=[x (n), x (n-1) ..., x (n-L+1)]T, wherein, L is represented
The length of sef-adapting filter, L=512 or 1024, subscript T represents transposition；
Using voice signal centrifugal pump x (n) as input signal input adaptive wave filter, the adaptive of current time n is obtained
Answer the output signal of wave filter Wherein, w (n) is the weights of current n moment sef-adapting filter
Vector, its length is equal to L, and initial value is null vector；
C, echo cancelltion
A is walked to obtained current time n desired signal centrifugal pump d (n), current time n sef-adapting filter is subtracted
Output signalObtain current time n error signal e (n)；And by current time n error signal e (n), as disappearing
Except the clean signal after echo sends distal end to,；
D, the error vector of sef-adapting filter are calculated
D1, the sef-adapting filter input vector x (n) with current time n to moment n-M+1, x (n-1) ..., x (n-M+
1) sef-adapting filter, is constituted in current time n input matrix X (n), X (n)=[x (n), x (n-1) ..., x (n-M+
1)], wherein, M be affine projection exponent number, its span be 4~10；
D2, the desired signal d (n) with current time n to moment n-M+1, d (n-1) ..., d (n-M+1), are constituted current
Moment n desired signal vector d (n), d (n)=[d (n), d (n-1) ..., d (n-M+1)]T；
D3, current time n desired signal vector d (n) subtracted current time n input matrix X (n) transposition with from
The weight vector w (n) of adaptive filter product, obtains the error vector e (n) of current time n sef-adapting filter, e (n)
=d (n)-XT(n)w(n)；
E, sef-adapting filter right value update
E1, scale factor calculating
I-th of scale factor g of current time n sef-adapting filteri(n), drawn by following formula：
Wherein, wi(n) it is i-th element of the sef-adapting filter in the weight vector w (n) at moment at current time n, | |
||1Represent to ask vectorial 1- norms, κ is proportionality control parameter κ ∈ [- 1,1], and ε is regularization parameter, and its value is
0.0001~0.001；
By all proportions factor g of current time n sef-adapting filter1(n),g2(n)...gi(n)...,gL(n), structure
Into sef-adapting filter current time n column vector G (n),
G (n)=[g1(n),...gi(n)...,gL(n)]T；
E2, proportional intermediary matrix calculating
By the scale factor column vector G (0) of initial time n=0 sef-adapting filter diagonal matrix diag { G (0) }
It is multiplied with the input matrix X (0) of initial time n=0 sef-adapting filter, obtains initial time n=0 proportional middle square
Battle array P (0), P (0)=diag { G (0) } X (0)；
According to sef-adapting filter in current time n scale factor column vector G (n) and the adaptive-filtering at current time
Device input vector x (n), the P (n) of the proportional intermediary matrix for the current time n for obtaining L rows M row is calculated by following formula,
P (n)=[G (n) ⊙ x (n), P-1(n)],
Wherein, ⊙ represents point multiplication operation, P-1(n) the M-1 row before the intermediary matrix P (n-1) of previous moment (n-1)
Vector is constituted；
E3, sef-adapting filter right value update
The right value update increment Delta w (n) of current time n sef-adapting filter, is drawn by following formula：
Δ w (n)=P (n) [δ × I+XT(n)P(n)]-1E (n),
Wherein, I represents M × M unit matrix, []-1Finding the inverse matrix is represented, δ is stability contorting parameter, its value
For 0.001~0.01；
F, the calculating for combining step-length
F1, the combination step-length auxiliary parameter a (n) at current time are drawn as follows：
A (n)=a (n-1)+μa(μ1-μ2)[λ(n-1)(1-λ(n-1))+θ]e(n)xT(n)Δw(n-1)
Wherein, μaIt is the study control parameter for combining step-length, value is 100~300；μ1It is big step-length, its value is 1；μ2It is
Small step-length, its span is 0 ＜ μ2≤0.1；θ is combination step-length steadiness parameter, and value is 1 × 10-5；Δ w (n-1) is previous
The right value update increment of moment (n-1) sef-adapting filter, its initial value Δ w (0) is 0；Auxiliary parameter a (n) initial value is a
(0)=4；λ (n-1) is the step-length hybrid parameter of previous moment (n-1), and its initial value λ (0) is 0.
F2, step-length hybrid parameter calculating
Current time n step-length hybrid parameter λ (n) is drawn by the following method：
IfThen λ (n)=0；
IfThen λ (n)=1；
IfThen
Wherein, C is a mixing control constant, and span is 2~4；
F3, the calculating for combining step-length
Current time n combination step size mu (n) is drawn by following formula：
μ (n)=λ (n) μ1+(1-λ(n))μ2
G, sef-adapting filter right value update
Calculate what is obtained according to the step E right value update increment Delta w (n) and step F for calculating obtained sef-adapting filter
Step size mu (n) is combined, the weight vector w (n+1), w (n+1)=w (n) of the sef-adapting filter for obtaining subsequent time (n+1) is updated
+μ(n)Δw(n)；
H, repetition
N=n+1 is made, the step of repeating A, B, C, D, E, F, G, until end of conversation.
Emulation experiment：
In order to verify effectiveness of the invention, emulation experiment has been carried out.
The impulse response of echo channel is formed in a closed room in emulation experiment, and its length is 512, sees Fig. 1.
The signal x (n) that telelecture is played back is real voice signal, and sample frequency is 8000Hz, sampled point number 40,
000, see Fig. 2.The echo signal gathered at telephone microphone, is shown in Fig. 3, the signal to noise ratio of measuring environment is 30dB.
According to above experiment condition, the memory of the invention from different step size mu=1 and μ=0.05 improves proportional affine throwing
Shadow algorithm (MIPAPA) carries out echo cancellor experiment.The identical parameters of three are selected as ε=0.001, κ=0, δ=0.01, imitated
Penetrate projection order M=4；The newly-increased parameter value of the present invention is big step size mu1=1, small step size mu2=0.05, μa=100, C=2, θ
=10-5。
Carry out the performance of comparison algorithm by using normalization imbalance, be defined asIt is single
Position is decibel (dB).Fig. 4 shows the normalization imbalance performance of algorithm.
Figure 4, it is seen that the present invention is 2 × 104Moment restrains, and step-length is the MIPAPA of μ=1 also 2 × 104When
Carve convergence, and step-length for μ=0.05 MIPAPA 4 × 104Moment restrains；Present invention obtains be μ=0.05 than step-length
The faster convergence rates of MIPAPA.The steady-state error of the present invention is -20dB, and step-length is the MIPAPA of μ=0.05 steady-state error
Also it is -20dB；And the steady-state error that step-length is the MIPAPA of μ=1 is -10dB, compared with MIPAPA of the step-length for μ=1, this hair
Bright steady-state error reduces 10dB.Therefore, the existing quick convergence rate of the present invention and but also with small steady-state error, its time
Sound effective value is more preferable.
Claims (1)
1. one kind combination proportional affine projection echo cancel method of step-length, its step is as follows：
A, signal acquisition
The signal sampling that the loudspeaker of phone is sent, obtains current time n voice signal centrifugal pump x (n)；Meanwhile, by electricity
The signal sampling that the microphone of words is induced, obtains current time n desired signal centrifugal pump d (n)；
B, echo signal estimation
Value x (n) by voice signal centrifugal pump x (n) at n to the n-L+1 moment, x (n-1) ..., x (n-L+1), when constituting current
Carve n sef-adapting filter input vector x (n), x (n)=[x (n), x (n-1) ..., x (n-L+1)]T, wherein, L represents adaptive
The length of wave filter is answered, L=512 or 1024, subscript T represents transposition；
Using voice signal centrifugal pump x (n) as input signal input adaptive wave filter, current time n adaptive filter is obtained
The output signal of ripple device Wherein, w (n) is the weight vector of current n moment sef-adapting filter,
Its length is equal to L, and initial value is null vector；
C, echo cancelltion
A is walked to obtained current time n desired signal centrifugal pump d (n), subtract current time n sef-adapting filter it is defeated
Go out signalObtain current time n error signal e (n)；And by current time n error signal e (n), as eliminating back
Clean signal after sound sends distal end to；
D, the error vector of sef-adapting filter are calculated
D1, the sef-adapting filter input vector x (n) with current time n to moment n-M+1, x (n-1) ..., x (n-M+1),
Sef-adapting filter is constituted in current time n input matrix X (n), X (n)=[x (n), x (n-1) ..., x (n-M+1)], its
In, M is affine projection exponent number, and its span is 4~10；
D2, the desired signal d (n) with current time n to moment n-M+1, d (n-1) ..., d (n-M+1), constitute current time n
Desired signal vector d (n), d (n)=[d (n), d (n-1) ..., d (n-M+1)]T；
D3, current time n desired signal vector d (n) subtracted current time n input matrix X (n) transposition with it is adaptive
The weight vector w (n) of wave filter product, obtains the error vector e (n) of current time n sef-adapting filter, e (n)=d
(n)-XT(n)w(n)；
E, sef-adapting filter right value update
E1, scale factor calculating
I-th of scale factor g of current time n sef-adapting filteri(n), drawn by following formula：
<mrow>
<msub>
<mi>g</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mn>1</mn>
<mo>-</mo>
<mi>&kappa;</mi>
</mrow>
<mrow>
<mn>2</mn>
<mi>L</mi>
</mrow>
</mfrac>
<mo>+</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>&kappa;</mi>
<mo>)</mo>
</mrow>
<mfrac>
<mrow>
<mo>|</mo>
<msub>
<mi>w</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>|</mo>
</mrow>
<mrow>
<mn>2</mn>
<mo>|</mo>
<mo>|</mo>
<mi>w</mi>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>1</mn>
</msub>
<mo>+</mo>
<mi>&epsiv;</mi>
</mrow>
</mfrac>
</mrow>
Wherein, wi(n) it is i-th element of the sef-adapting filter in the weight vector w (n) at moment at current time n, | | | |1Table
Show ask vector 1- norms, κ is proportionality control parameter κ ∈ [- 1,1], and ε is regularization parameter, its value be 0.0001~
0.001；
By all proportions factor g of current time n sef-adapting filter1(n),g2(n)...gi(n)...,gL(n), constitute certainly
Adaptive filter current time n column vector G (n),
G (n)=[g1(n),...gi(n)...,gL(n)]T；
E2, proportional intermediary matrix calculating
By the scale factor column vector G (0) of initial time n=0 sef-adapting filter diagonal matrix
Diag { G (0) } is multiplied with the input matrix X (0) of initial time n=0 sef-adapting filter, obtains initial time n=0
Proportional intermediary matrix P (0), P (0)=diag { G (0) } X (0)；
It is defeated in current time n scale factor column vector G (n) and the sef-adapting filter at current time according to sef-adapting filter
Incoming vector x (n), the P (n) of the proportional intermediary matrix for the current time n for obtaining L rows M row is calculated by following formula,
P (n)=[G (n) ⊙ x (n), P-1(n)],
Wherein, ⊙ represents point multiplication operation, P-1(n) the M-1 column vector before the intermediary matrix P (n-1) of previous moment (n-1)
Constitute；
E3, sef-adapting filter right value update
The right value update increment Delta w (n) of current time n sef-adapting filter, is drawn by following formula：
Δ w (n)=P (n) [δ × I+XT(n)P(n)]-1E (n),
Wherein, I represents M × M unit matrix, []-1Finding the inverse matrix is represented, δ is that stability contorting parameter, its value are
0.001~0.01；
F, the calculating for combining step-length
F1, the combination step-length auxiliary parameter a (n) at current time are drawn as follows：
A (n)=a (n-1)+μa(μ1-μ2)[λ(n-1)(1-λ(n-1))+θ]e(n)xT(n)Δw(n-1)
Wherein, μaIt is the study control parameter for combining step-length, value is 100~300；μ1It is big step-length, its value is 1；μ2It is small step
Long, its span is 0 ＜ μ2≤0.1；θ is combination step-length steadiness parameter, and value is 1 × 10-5；Δ w (n-1) is previous moment
(n-1) the right value update increment of sef-adapting filter, its initial value Δ w (0) is 0；Auxiliary parameter a (n) initial value is a (0)
=4；λ (n-1) is the step-length hybrid parameter of previous moment (n-1), and its initial value λ (0) is 0；
F2, step-length hybrid parameter calculating
Current time n step-length hybrid parameter λ (n) is drawn by the following method：
IfThen λ (n)=0；
IfThen λ (n)=1；
IfThen
Wherein, C is a mixing control constant, and span is 2~4；
F3, the calculating for combining step-length
Current time n combination step size mu (n) is drawn by following formula：
μ (n)=λ (n) μ1+(1-λ(n))μ2
G, sef-adapting filter right value update
Obtained combination is calculated according to the step E right value update increment Delta w (n) and step F for calculating obtained sef-adapting filter
Step size mu (n), updates the weight vector w (n+1), w (n+1)=w (n)+μ for the sef-adapting filter for obtaining subsequent time (n+1)
(n)Δw(n)；
H, repetition
N=n+1 is made, the step of repeating A, B, C, D, E, F, G, until end of conversation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710155109.6A CN107105111B (en) | 2017-03-15 | 2017-03-15 | A kind of proportional affine projection echo cancel method of combination step-length |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710155109.6A CN107105111B (en) | 2017-03-15 | 2017-03-15 | A kind of proportional affine projection echo cancel method of combination step-length |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107105111A true CN107105111A (en) | 2017-08-29 |
| CN107105111B CN107105111B (en) | 2019-08-02 |
Family
ID=59675595
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710155109.6A Expired - Fee Related CN107105111B (en) | 2017-03-15 | 2017-03-15 | A kind of proportional affine projection echo cancel method of combination step-length |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107105111B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107819963A (en) * | 2017-09-13 | 2018-03-20 | 西南交通大学 | A kind of minimum of convex combination cube absolute value echo cancel method |
| CN108877824A (en) * | 2018-05-31 | 2018-11-23 | 西南交通大学 | A kind of combination step-length echo cancel method that tracking performance is high |
| CN109003620A (en) * | 2018-05-24 | 2018-12-14 | 北京潘达互娱科技有限公司 | A kind of echo removing method, device, electronic equipment and storage medium |
| CN111868826A (en) * | 2018-12-17 | 2020-10-30 | 深圳市汇顶科技股份有限公司 | Adaptive filtering method, device, equipment and storage medium in echo cancellation |
| CN113364380A (en) * | 2021-08-10 | 2021-09-07 | 成都爱旗科技有限公司 | Motor parameter determination method and device |
| CN117896467A (en) * | 2024-03-14 | 2024-04-16 | 苏州大学 | Echo cancellation method and system for stereo telephone communication |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101878637A (en) * | 2007-11-29 | 2010-11-03 | 艾利森电话股份有限公司 | A method and arrangement for echo cancellation of voice signals |
| US8254588B2 (en) * | 2007-11-13 | 2012-08-28 | Stmicroelectronics Asia Pacific Pte., Ltd. | System and method for providing step size control for subband affine projection filters for echo cancellation applications |
| CN104050971A (en) * | 2013-03-15 | 2014-09-17 | 杜比实验室特许公司 | Acoustic echo mitigating apparatus and method, audio processing apparatus, and voice communication terminal |
| CN104270539A (en) * | 2014-09-12 | 2015-01-07 | 西南交通大学 | Proportional affine projection echo elimination method based on coefficient difference |
| CN104683614A (en) * | 2015-03-24 | 2015-06-03 | 西南交通大学 | Adaptive echo cancellation method adopting memory proportionate affine projection and based on M-estimation |
| CN104601837B (en) * | 2014-12-22 | 2016-03-02 | 西南交通大学 | A kind of robust convex combination self adaptation listener's echo removing method |
| CN105721729A (en) * | 2016-04-26 | 2016-06-29 | 西南交通大学 | Echo cancellation method based on block-sparse proportionate reuse weight coefficient affine projection |
| CN106448695A (en) * | 2016-09-28 | 2017-02-22 | 天津大学 | Double-end call robust variable-order variable-step-size affine projection method |
-
2017
- 2017-03-15 CN CN201710155109.6A patent/CN107105111B/en not_active Expired - Fee Related
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8254588B2 (en) * | 2007-11-13 | 2012-08-28 | Stmicroelectronics Asia Pacific Pte., Ltd. | System and method for providing step size control for subband affine projection filters for echo cancellation applications |
| CN101878637A (en) * | 2007-11-29 | 2010-11-03 | 艾利森电话股份有限公司 | A method and arrangement for echo cancellation of voice signals |
| CN104050971A (en) * | 2013-03-15 | 2014-09-17 | 杜比实验室特许公司 | Acoustic echo mitigating apparatus and method, audio processing apparatus, and voice communication terminal |
| CN104270539A (en) * | 2014-09-12 | 2015-01-07 | 西南交通大学 | Proportional affine projection echo elimination method based on coefficient difference |
| CN104601837B (en) * | 2014-12-22 | 2016-03-02 | 西南交通大学 | A kind of robust convex combination self adaptation listener's echo removing method |
| CN104683614A (en) * | 2015-03-24 | 2015-06-03 | 西南交通大学 | Adaptive echo cancellation method adopting memory proportionate affine projection and based on M-estimation |
| CN105721729A (en) * | 2016-04-26 | 2016-06-29 | 西南交通大学 | Echo cancellation method based on block-sparse proportionate reuse weight coefficient affine projection |
| CN106448695A (en) * | 2016-09-28 | 2017-02-22 | 天津大学 | Double-end call robust variable-order variable-step-size affine projection method |
Non-Patent Citations (2)
| Title |
|---|
| YI YU,HAIQUAN ZHAO,BAODONG CHEN: "Sparseness-controlled Proportionate Affine Projection Sign Algorithm for Acoustic Echo Cancellation", 《CIRCUITS SYST SIGNAL PROCESS》 * |
| 文昊翔, 陈隆道, 蔡忠法: "改进仿射投影算法及其在电子回声消除中的应用", 《电子学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107819963A (en) * | 2017-09-13 | 2018-03-20 | 西南交通大学 | A kind of minimum of convex combination cube absolute value echo cancel method |
| CN109003620A (en) * | 2018-05-24 | 2018-12-14 | 北京潘达互娱科技有限公司 | A kind of echo removing method, device, electronic equipment and storage medium |
| CN108877824A (en) * | 2018-05-31 | 2018-11-23 | 西南交通大学 | A kind of combination step-length echo cancel method that tracking performance is high |
| CN111868826A (en) * | 2018-12-17 | 2020-10-30 | 深圳市汇顶科技股份有限公司 | Adaptive filtering method, device, equipment and storage medium in echo cancellation |
| CN113364380A (en) * | 2021-08-10 | 2021-09-07 | 成都爱旗科技有限公司 | Motor parameter determination method and device |
| CN117896467A (en) * | 2024-03-14 | 2024-04-16 | 苏州大学 | Echo cancellation method and system for stereo telephone communication |
| CN117896467B (en) * | 2024-03-14 | 2024-05-31 | 苏州大学 | Echo cancellation method and system for stereo telephone communication |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107105111B (en) | 2019-08-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107105111B (en) | A kind of proportional affine projection echo cancel method of combination step-length | |
| CN111768796B (en) | Acoustic echo cancellation and dereverberation method and device | |
| CN103369162B (en) | A kind of listener's echo self adaptive elimination method of low complex degree | |
| CN104883462B (en) | A kind of sef-adapting filter and filtering method for eliminating acoustic echo | |
| CN104683614B (en) | Based on the proportional illumination-imitation projection self-adoptive echo cancel method of memory that M estimates | |
| CN104601837B (en) | A kind of robust convex combination self adaptation listener's echo removing method | |
| CN104410761B (en) | A kind of affine projection symbol subband convex combination adaptive echo cancellation method | |
| CN104270539B (en) | Based on the proportional affine projection echo cancel method of coefficient difference | |
| CN105070296A (en) | Active factor set membership proportional sub band self-adaption echo cancellation method | |
| CN105721729B (en) | Based on the sparse proportional echo cancel method for reusing weight coefficient affine projection of block | |
| CN106210370A (en) | A kind of affine projection maximum entropy sub-band adaptive echo cancel method | |
| CN104506746A (en) | Improved convex combination decorrelation proportionate self-adaption echo cancellation method | |
| CN107333017B (en) | A kind of reuse weight coefficient exponential type error adaptive echo cancellation method | |
| CN104144269B (en) | A kind of proportional self adaptation listener's echo removing method based on decorrelation | |
| CN104158512B (en) | The adaptive sparse system identifying method that a kind of shock resistance based on the independent activities factor is disturbed | |
| CN108877824B (en) | Combined step echo cancellation method with high tracking performance | |
| Buchris et al. | On the design of time-domain differential microphone arrays | |
| CN104410762B (en) | Sane echo cancelltion method in hands-free speaking system | |
| CN103929704B (en) | The method and system that a kind of adaptive acoustic feedback based on transform domain is eliminated | |
| CN105070295A (en) | Adaptive method, applied to echo cancellation, of active factor proportional sub band | |
| CN105788605B (en) | A kind of affine projection symbol echo cancel method based on orthogonal transformation | |
| CN106157965B (en) | A kind of zero norm collection person's illumination-imitation projection self-adoptive echo cancel method reused based on weight vector | |
| CN109935238B (en) | Echo cancellation method, device and terminal equipment | |
| CN106060295B (en) | A kind of proportional affine projection echo cancel method of convex combination coefficient difference | |
| CN112802487B (en) | Echo processing method, device and system |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190802 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |