CN106412352B - A kind of time-frequency memory subband ratio adaptive echo cancellation method - Google Patents
A kind of time-frequency memory subband ratio adaptive echo cancellation method Download PDFInfo
- Publication number
- CN106412352B CN106412352B CN201610832986.8A CN201610832986A CN106412352B CN 106412352 B CN106412352 B CN 106412352B CN 201610832986 A CN201610832986 A CN 201610832986A CN 106412352 B CN106412352 B CN 106412352B
- Authority
- CN
- China
- Prior art keywords
- subband
- echo
- vector
- signal
- moment
- 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
- 238000000034 method Methods 0.000 title claims abstract description 47
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 32
- 238000004458 analytical method Methods 0.000 claims description 17
- 239000011159 matrix material Substances 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000008030 elimination Effects 0.000 claims description 6
- 238000003379 elimination reaction Methods 0.000 claims description 6
- 230000011664 signaling Effects 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 230000017105 transposition Effects 0.000 claims description 2
- 238000005192 partition Methods 0.000 claims 1
- 238000004422 calculation algorithm Methods 0.000 abstract description 20
- 238000005516 engineering process Methods 0.000 abstract description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 9
- 238000013459 approach Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 238000001914 filtration Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000010195 expression analysis Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000001228 spectrum 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
- H04M9/082—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
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
- G10L21/0232—Processing in the frequency domain
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- 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)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
- Filters That Use Time-Delay Elements (AREA)
Abstract
The invention discloses a kind of time-frequencies to remember subband ratio adaptive echo cancellation method, belongs to electroacoustic adaptive echo technology for eliminating field.For Echo Cancellation Problem, a kind of time-frequency memory subband ratio adaptive echo cancellation method is provided.On the one hand the invention is directed to the non-stationary property of input echo signal, using sub-band structure, carry out decorrelation to input echo signal on frequency domain, carry out prewhitening from time domain using AP algorithm, obtain preferable echo cancellation performance；On the other hand using the scale factor with time-frequency memory, the sparsity of echo channel is resisted, faster rate of convergence is obtained.
Description
Technical field
Present invention relates particularly to a kind of time-frequencies to remember subband ratio adaptive echo cancellation method, belongs to electroacoustic and adaptively returns
Sound technology for eliminating field.
Background technique
In hands-free phone meeting or mobile phone communication, due to call environment, it can often be heard from loudspeaker certainly
The sound of oneself delay, the sound of the delay are known as echo.Elimination such echo improves speech quality, usually in sound system
Inside installation echo cancelling device (acoustic echo canceller, AEC) generates and phase phase identical as echo amplitude
Anti- acoustic signal, and signal will be received in sound receiving end and be added with this with the signal of width antiphase, so that it is dry to eliminate echo
It disturbs.
The key problem of echo cancellor is the tracking of echo channel and the estimation of echo, is ideally generated as AEC
Echo estimated value is identical as echo amplitude, and when opposite in phase, the echo in system can be suppressed completely.Currently, echo channel
Tracking and echo estimation generally use auto-adaptive filtering technique, design various filters (FIR filter, iir filter and subband
Filter etc.) external echo channel is simulated, and the tap of filter is updated in real time using adaptive algorithm.
Therefore, the adaptive updates rate of convergence and accuracy of filter are to influence two principal elements of echo cancellation performance.Filtering
Device rate of convergence is too low, and the decline of echo estimated value tracking ability causes rejection to reduce；Filter accuracy of estimation is too low,
Residual echo is larger when leading to stable state, influences normally to converse.
Voice signal is a kind of non-stationary signal, its time-frequency domain characteristic can be therefore traditional adaptive with time change
Algorithm is answered, such as lowest mean square (least mean square, LMS) algorithm, normalization minimum mean-square (Normalized least
Mean square, NLMS) algorithm is under voice signal background, and rate of convergence and stable state accurate performance are all by huge limit
System；In addition, another feature of echo cancellor is the sparse characteristic of its echo channel, i.e. the energy of echo channel is concentrated mainly on
On a small number of significant tap weight coefficients, all close to zero, this sparsity is serious to be constrained a large amount of tap weights coefficient amplitude
The convergence dynamic property of traditional adaptive algorithm passes through the inspection to existing technical literature for both the above Echo cancellation feature
Rope, currently used method have several:
Bibliography 1: in Chinese invention patent application number 201510028006.4, a kind of entitled " improved convex combination
In the proportional adaptive echo cancellation method of decorrelation ", it is directed to the sparsity of echo channel, using the convex combination knot of filter
Structure effectively alleviates channel sparsity to algorithmic statement rate in conjunction with decorrelation adaptive algorithm and ratio adaptive algorithm
Influence, obtain faster rate of convergence, low steady-state error and preferable interference free performance, but due to needing two sets to filter
The composition convex combination structure of wave device, expends resource and computation complexity is all larger, and the characteristic of algorithm, which mixes convex combination, joins
Number etc. is more sensitive, increases hard-wired difficulty.
Bibliography 2: " An improved multiband-structured subband adaptive filter
Algorithm " (Chinese name: a kind of improved more included structon band adaptive filter algorithm Yang, F., Wu, M., Ji, P.,
Yang, J., IEEE Signal Process.Lett., 2012,19, (10), pp.647-650) for the non-flat of echo signal
Steady characteristic, the method divided using subband, is divided into multiple subbands for the frequency spectrum of signal, so that each subband signal has white noise
The characteristic of sound reduces its non-flat stability, in addition, using affine projection algorithm (Affine projection, AP) to every height
It brings line filter tap weights coefficient update into, so that entire more new construction has time-frequency two-dimensional characteristic, effectively raises echo
The performance of inhibition, but its there is no consider be echo channel sparse characteristic, and the more situation of subband number can be brought
Huge computation complexity.
Bibliography 3: " An efficient proportionate affine projection algorithm for
Echo cancellation " (Chinese name: effective Echo cancellation affine projection algorithm Paleologu, C., a Ciochina,
S., Benesty, J.IEEE Signal Process.Lett., 2010,17, (2), pp.165-168) for echo signal
Sparse characteristic and echo it is non-stationary, updated using filter tap weights coefficient ratio and in AP algorithm in the white of time domain
Change effect, improves the rate of convergence of algorithm and the elimination performance of echo, but this method is not due to considering non-stationary signal
Frequency domain characteristic, there are still the problems that rate of convergence is slow.
Summary of the invention
The present invention is directed to above-mentioned Echo Cancellation Problem, provides a kind of time-frequency memory subband ratio adaptive echo elimination side
Method.Method specifically includes the following steps:
Step 1
Remote signaling x (n) and proximal end the signal d (n) comprising echo received are input to analysis filter group F0
(z), F1(z) ... FN-1(z) in, wherein n indicates time scale, Fi(z) domain Z of i=1,2 ... N-1 expression analysis filter group
Transmission function, analysis filter group include N number of analysis filter, and the length of analysis filter is M, on frequency domain uniformly by signal
It is divided into the subband remote signaling x of N number of equiband0(n), x1(n) ... xN-1(n) and the subband near end signal d of N number of equiband0
(n), d1(n) ... dN-1(n)；
Step 2
Distal end subband signal x0(n), x1(n) ... xN-1(n) and near end signal subband d0(n), d1(n) ... dN-1(n) it carries out
N times for reducing rate extracts, the subband signal x after obtaining reduction of speed0(p), x1(p) ... xN-1(p)、d0(p), d1(p) ... dN-1
(p), wherein p indicates time scale, and p=n/N；
Step 3
P moment N number of sub-band echo valuation vector is calculated by formula one,
Wherein Xi(p)=[xi(p), xi(p-1) ... xi(p-D+1)] be i-th of input subband matrix, dimension be L ×
D, D are projection order, xi(p)=[xi(pN), xi(pN-1) ..., xi(pN-L+1)]TFor i-th of subband input vector of filter,
W (p)=[w0(p), w1(p) ... wL-1(p)]TFor p moment filter tap weight vector, specific value is by p-1 moment echo
It is known that and originating 0 moment, w (0)=[0,0 ... 0] during eliminationT, L is tap number, and L=256~1024, T are indicated
Conjugate transposition operation；
Step 4
The estimation error vector of i-th of subband is calculated by formula two,
Wherein eI, D(p)=[eI, 0(pN), eI, 1(pN-1) ... eI, L-1(pN-L+1)]TIt is the error subband that dimension is L × 1
Vector, dI, D(p)=[dI, 0(pN), dI, 1(pN-1) ... dI, L-1(pN-L+1)]TIndicate proximal end subband signal vector；
Step 5
I-th of subband scale factor vector is calculated by formula three,
ci(p)=[cI, 0(p), cI, 1(p) ... cI, L-1(p)]TFormula three
cI, lIt (p) is ci(p) i-th of element,
Wherein ζ takes 0.5 as the ratio scale factor, and -1 < ζ < 1 of value range, σ=0.001 is to format the factor, θI-1, l
(p) indicate that (i-1)-th subband signal is the frequency domain memory fact of scale factor to the increment contribution factor of first of weight coefficient, when
When i=0, θI-1, l(p)=0 p-1 moment, p-2 moment ... ..., the subband ratio at p-D+1 moment, while according to the method described above, are calculated
Example is because of subvector ci(p-1), ci(p-2) ... ..., ci(p-D+1), D is that the time domain of scale factor remembers scale；
Step 6
Ratio input matrix is calculated in formula four by subband input signal matrix and scale factor vector
WhereinIndicate the hadmard product of vector；
Step 7
I-th of subband weight vector increment θ is calculated by formula fivei(p)
Wherein λ=0.001 is to format the factor, and I is that D × D ties up unit matrix, and λ I is the diagonal matrix that diagonal element is λ,
The stability that the purpose is to guarantee to calculate, μ=0.1 are step factor, 0 < μ < 2 of value range；
Step 8
Step 4 is returned to, all subband weight vector increments, i.e. θ are calculated according to step 4 to step 7i(p) i=1,
2,…N-1；
Step 9
Tap weight value vector is updated according to formula six,
The weight vector w (p+1) at p+1 moment is finally obtained, dimension is L × 1；
Step 10
Step 3 is returned to, p+1 moment N number of sub-band echo is calculated and estimates vector,
Wherein XiIt (p+1) is i-th of subband input matrix of p+1 moment filter, dimension is L × D, and w (p+1) is to calculate in step 9
P+1 moment filter tap weight vector out, dimension are L × 1；
Step 11
D is inputted from the moment proximal end p+1 for including echo-signal by formula sevenI, D(p+1) it is subtracted in (i=1,2 ... N-1)
Echo interference estimated value,
Wherein eI, D(p+1) it is the error signal interfered after eliminating, includes proximal end useful signal and residual echo, dimension
For L × 1, dI, D(p+1)=[dI, 0(pN+1), dI, 1(pN) ... dI, L-1(pN-L)]TIndicate p+1 moment proximal end subband signal,
Dimension is L × 1,
It is finally completed echo interference elimination.
Further, N number of analysis filter in the step 1 in analysis filter group is discrete cosine filter.
Further, D value range is 4-8 in the step 3.
The beneficial effects of the present invention are: time-frequency memory subband ratio adaptive echo cancellation method of the invention and feedback
Method uses sub-band structure for the non-stationary property of echo signal on frequency domain, by proximal end input and far end-echo signal point
Multiple subbands are segmented into, the correlation of signal is effectively reduced；On the other hand, take multidimensional projection input composition defeated in the time domain
Enter matrix, participate in filter tap weight vector and update operation, plays the role of time-frequency two-dimensional prewhitening simultaneously.
The characteristics of for echo channel sparsity, the invention proposes time-frequency two-dimensionals to remember ratio adaptive approach, is counting
When calculating scale factor, it is contemplated that time memory property, i.e. time delay D and frequency domain Memorability, i.e. the weight coefficient increment of different sub-band
Contribution factor θI, l(p), the rate of convergence and steady-state performance of adaptive algorithm are improved.
The present invention provides a unified framework for all proportions adaptive echo cancellation method, that is, different ginsengs is arranged
Number, so that it may obtain different ratio adaptive approach, such as work as N=1, illustrate to only exist only one height in echo cancelling system
Band, the method degenerate into improved memory sex ratio affine projection (Memory improved Proportionate affine
Projection algorithm, MIPAPA) adaptive approach, so as to which different parameters is arranged according to the actual situation, selection
Different echo cancellor frameworks and method.
Detailed description of the invention
Fig. 1 is the sub-band structure figure of the method for the present invention in embodiment.
Fig. 2 is the analysis filter group structure chart of the method for the present invention in embodiment.
Fig. 3 is the condition of sparse channel schematic diagram of the method for the present invention in embodiment.
Fig. 4 is that 2 method of 5dB low signal-to-noise ratio environment Publication about Document, 3 method of document and weight coefficient normalization imbalance of the invention are bent
Line comparison schematic diagram.
Fig. 5 is that 2 method of 20dB high s/n ratio environment Publication about Document, 3 method of document and weight coefficient of the invention normalize imbalance
Curve comparison schematic diagram.
Specific embodiment
Description of specific embodiments of the present invention with reference to the accompanying drawing:
The present invention is directed to Echo Cancellation Problem, provides a kind of time-frequency memory subband ratio adaptive echo cancellation method.
On the one hand the invention is directed to the non-stationary property of input echo signal, using sub-band structure, to input echo signal on frequency domain
Decorrelation is carried out, prewhitening is carried out from time domain using AP algorithm, obtains preferable echo cancellation performance；On the other hand band is utilized
The scale factor for having time-frequency to remember, resists the sparsity of echo channel, obtains faster rate of convergence.
Verify helpfulness of the invention using a simulation example, and with existing literature 2 in background technique and document 3
Method is made comparisons.
Using technical solution of the present invention, sub-band structure is as shown in Figure 1, the analysis filter group structure such as Fig. 2 used
It is shown.N number of analysis filter in analysis filter group is discrete cosine filter.
The standardized sparse channel that echo channel uses international telecommunication standard ITU-TG168 to define, as shown in figure 3, channel is cut
Take length L=512.Remote end input signal x (n) be 10 rank autoregression AR (10) signals, autoregressive coefficient be (5.3217 ,-
9.2948,7.0933, -2.8125,2.5805, -2.4230,0.3747,2.2628, -0.3028,1.7444,1.1053),
With high-intensitive correlation properties.Signal-to-noise ratio (SNR) is set as 5dB or 20dB, is respectively compared various methods in low signal-to-noise ratio and height
Performance under the conditions of signal-to-noise ratio.
The echo signal d (n) for picking up proximal end according to sample frequency 8k in proximal end, samples 8000 samples, wherein d (n) altogether
It simultaneously include echo signal, remote signaling x (n) and noise.
All parameters chosen in emulation experiment are as shown in table 1.
Table 1
Fig. 4 show 2 method of 5dB low signal-to-noise ratio environment Publication about Document, 3 method of document and weight coefficient normalization of the invention and loses
Adjust curve comparison schematic diagram, wherein curve 1 represents the present invention, and curve 2 represents document 2, and curve 3 represents document 3, it can be seen that
It restrains under step factor unanimous circumstances, faster, steady-state performance is more excellent for rate of convergence of the present invention.
Fig. 5 show 2 method of 20dB high s/n ratio environment Publication about Document, 3 method of document and weight coefficient of the invention and normalizes
Imbalance curve compares, wherein curve 4 represents the present invention, and curve 5 represents document 2, and curve 6 represents document 3, it can be seen that believes in height
It makes an uproar than under, the present invention still obtains faster rate of convergence and preferable steady-state performance.
Preferable improved technical solution is as follows:
The present invention is not limited to that it is sparse following scale factor improvement channel can also be chosen using above-mentioned memory scale factor
Property
μ μ-law scale factor
Wherein βi,l(p) calculation is
βi,l(p)
=max ρ max [δ, T (| w0(p) |) ..., T (| wl-1(p) |), T (| wl(p)|)]}
T () function is defined as
T(|wl(p) |) and=ln (1+ η | wl(p) |), the ε of η=1/
Wherein ρ, δ and ε are the positive number of very little, and typical choosing value is ρ=5/L, and δ=0.01, ε are according to ambient noise, choosing
It is taken as the order of magnitude of noise.
Modified scale factor
Wherein scale factor ζ is variable, and calculation is
Γ × max (R (p)) is choosing value thresholding, and the representative value of Γ is 0.1.κ1Typical value range be -1~-0.98, κ2
Typical value range be 0.98~1.Vector R(p)Calculation is
R (p)=[r0(p),r1(p),…rL-1(p)]
rl(p)=max ρ max [| w0(p)|,…,|wL-1(p)|],|wl(p)|}
The present invention is according to subband number N, affine projection dimension D and memory scale factor δi,l(k) setting, does not limit to
In sub-band adaptive filtering；Work as N=1, illustrates to only exist only one subband in echo cancelling system, the method, which degenerates into, to be changed
Into memory sex ratio affine projection (Memory improved Proportionate affine projection
Algorithm, MIPAPA) adaptive approach；Work as D=1, δi,l(k)=0 when, illustrate that the dimension inputted in echo cancelling system is
1 dimension, the method can regard improved subband ratio adaptive approach as；Work as D=1, N=1 and δi,l(k)=0 when, the method is degraded
For improve subband normalization minimum mean-square (Improved Proportionate Normalized least mean square,
IPNLMS) adaptive approach.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, without departing from the principles of the present invention, it can also make several improvements and retouch, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (5)
1. a kind of time-frequency remembers subband ratio adaptive echo cancellation method, which is characterized in that the described method comprises the following steps:
Step 1
Remote signaling x (n) and proximal end the signal d (n) comprising echo received are input to analysis filter group F0(z),F1
(z),…FN-1(z) in, wherein n indicates time scale, Fi(z) i=1,2 ... N-1 indicate the domain the Z transmitting of analysis filter group
Function, analysis filter group include N number of analysis filter, and the length of analysis filter is M, by signal on frequency domain even partition
For the subband remote signaling x of N number of equiband0(n),x1(n),…xN-1(n) and the subband near end signal d of N number of equiband0(n),d1
(n),…dN-1(n)；
Step 2
Distal end subband signal x0(n),x1(n),…xN-1(n) and near end signal subband d0(n),d1(n),…dN-1(n) it is reduced
N times of rate extracts, the subband signal x after obtaining reduction of speed0(p),x1(p),…xN-1(p)、d0(p),d1(p),…dN-1(p),
Middle p indicates time scale, and p=n/N；
Step 3
P moment N number of sub-band echo valuation vector is calculated by formula one,
Wherein Xi(p)=[xi(p),xi(p-1),…xiIt (p-D+1)] is i-th of input subband matrix, dimension is L × D, and D is
The projection order of input subband matrix, xi(p)=[xi(pN),xi(pN-1),…,xi(pN-L+1)]TFor i-th of subband of filter
Input vector, w (p)=[w0(p),w1(p),…wL-1(p)]TFor p moment filter tap weight vector, specific value is by p-
It is known that and originating 0 moment, w (0)=[0,0 ... 0] in 1 moment echo cancellation processT, L be tap number, L=256~
1024, T indicate conjugate transposition operation；
Step 4
The estimation error vector of i-th of subband is calculated by formula two,
Wherein ei,D(p)=[ei,0(pN),ei,1(pN-1),…ei,L-1(pN-L+1)]TBe dimension be L × 1 error subband to
Amount, di,D(p)=[di,0(pN),di,1(pN-1),…di,L-1(pN-L+1)]TIndicate proximal end subband signal vector；
Step 5
I-th of subband scale factor vector is calculated by formula three,
ci(p)=[ci,0(p),ci,1(p),…ci,L-1(p)]TFormula three
ci,lIt (p) is ci(p) first of element, l=0,1,2 ... L-1,
Wherein ζ is the ratio scale factor, and -1 < ζ < 1 of value range, σ=0.001 is to format the factor, θi-1,l(p) the is indicated
I-1 subband signal is the frequency domain memory fact of scale factor to the increment contribution factor of first of weight coefficient, as i=0,
θi-1,l(p)=0 p-1 moment, p-2 moment ... ..., p-D, while according to the method described above, are calculated0The subband scale factor at+1 moment
Vector ci(p-1), ci(p-2) ... ..., ci(p-D0+ 1), D0For scale factor time domain remember scale, numerically with input
The projection order D of subband matrix is equal；
Step 6
Ratio input matrix is calculated in formula four by input subband matrix and scale factor vector
WhereinIndicate the hadmard product of vector；
Step 7
I-th of subband weight vector increment θ is calculated by formula fivei(p)
Wherein λ=0.001 is to format the factor, and I is that D × D ties up unit matrix, and λ I is the diagonal matrix that diagonal element is λ, and μ is
Step factor, 0 < μ < 2 of value range；
Step 8
Step 4 is returned to, all subband weight vector increments, i.e. θ are calculated according to step 4 to step 7i(p), i=1,2 ... N-
1；
Step 9
Tap weight value vector is updated according to formula six,
The weight vector w (p+1) at p+1 moment is finally obtained, dimension is L × 1；
Step 10
Step 3 is returned to, p+1 moment N number of sub-band echo is calculated and estimates vector,Wherein
XiIt (p+1) is i-th of subband input matrix of p+1 moment filter, dimension is L × D, and w (p+1) is calculated in step 9
P+1 moment filter tap weight vector, dimension are L × 1；
Step 11
D is inputted from the moment proximal end p+1 for including echo-signal by formula seveni,D(p+1) echo is subtracted in (i=1,2 ... N-1)
Interference estimate,
Wherein ei,D(p+1) be interference eliminate after error signal, include proximal end useful signal and residual echo, dimension be L ×
1, di,D(p+1)=[di,0(pN+1),di,1(pN),…di,L-1(pN-L)]TIndicate that p+1 moment proximal end subband signal, dimension are
L × 1,
It is finally completed echo interference elimination.
2. time-frequency as described in claim 1 remembers subband ratio adaptive echo cancellation method, which is characterized in that the step
N number of analysis filter in one in analysis filter group is discrete cosine filter.
3. time-frequency as described in claim 1 remembers subband ratio adaptive echo cancellation method, which is characterized in that the step
D value range is 4-8, and D in step 5 in three0=D.
4. time-frequency as described in claim 1 remembers subband ratio adaptive echo cancellation method, which is characterized in that the step
Scale factor is replaced using μ μ-law scale factor in five, is calculated by formula eight,
Wherein βi,l(p) calculation is
βi,l(p)=max ρ max [δ, T (| w0(p) |) ..., T (| wl-1(p) |), T (| wl(p)|)]}
T () function is defined as
T(|wl(p) |) and=ln (1+ η | wl(p) |), the ε of η=1/
Wherein ρ, δ and ε choosing value are ρ=5/L, and δ=0.01, ε are chosen for the order of magnitude of noise according to ambient noise.
5. time-frequency as described in claim 1 remembers subband ratio adaptive echo cancellation method, which is characterized in that the step
Scale factor is replaced using modified scale factor in five, is calculated by formula nine,
Wherein scale factor ζl(p) it can be changed, calculation is
Γ × max (R (p)) is choosing value thresholding, and the representative value of Γ is 0.1, κ1Value range be -1~-0.98, κ2Value model
Enclose is 0.98~1.Vector R (p) calculation is
R (p)=[r0(p),r1(p),…rL-1(p)]
rl(p)=max ρ max [| w0(p)|,…,|wL-1(p)|],|wl(p)|}。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610832986.8A CN106412352B (en) | 2016-09-19 | 2016-09-19 | A kind of time-frequency memory subband ratio adaptive echo cancellation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610832986.8A CN106412352B (en) | 2016-09-19 | 2016-09-19 | A kind of time-frequency memory subband ratio adaptive echo cancellation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106412352A CN106412352A (en) | 2017-02-15 |
| CN106412352B true CN106412352B (en) | 2019-06-14 |
Family
ID=57997848
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610832986.8A Active CN106412352B (en) | 2016-09-19 | 2016-09-19 | A kind of time-frequency memory subband ratio adaptive echo cancellation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106412352B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109412640B (en) * | 2018-11-22 | 2020-06-16 | 东南大学 | Nonlinear digital self-interference elimination device and method applied to full-duplex communication transceiver |
| CN109379501B (en) * | 2018-12-17 | 2021-12-21 | 嘉楠明芯(北京)科技有限公司 | Filtering method, device, equipment and medium for echo cancellation |
| US11770735B2 (en) | 2019-01-17 | 2023-09-26 | Nokia Technologies Oy | Overhead reduction in channel state information feedback |
| CN112542176B (en) * | 2020-11-04 | 2023-07-21 | 北京百度网讯科技有限公司 | Signal enhancement method, device and storage medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104506746B (en) * | 2015-01-20 | 2016-03-02 | 西南交通大学 | A kind of proportional adaptive echo cancellation method of convex combination decorrelation of improvement |
| CN105070296A (en) * | 2015-07-10 | 2015-11-18 | 西南交通大学 | Active factor set membership proportional sub band self-adaption echo cancellation method |
-
2016
- 2016-09-19 CN CN201610832986.8A patent/CN106412352B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106412352A (en) | 2017-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106412352B (en) | A kind of time-frequency memory subband ratio adaptive echo cancellation method | |
| TWI692754B (en) | Method and device for continuously updating coefficient vector of finite impulse response filter | |
| CN109727604A (en) | Frequency domain echo cancel method and computer storage media for speech recognition front-ends | |
| CN103369162B (en) | A kind of listener's echo self adaptive elimination method of low complex degree | |
| Fazel et al. | Deep Multitask Acoustic Echo Cancellation. | |
| CN104410761B (en) | A kind of affine projection symbol subband convex combination adaptive echo cancellation method | |
| CN104506746B (en) | A kind of proportional adaptive echo cancellation method of convex combination decorrelation of improvement | |
| CN109102794A (en) | M based on convex combination estimates the echo cancel method of proportional class affine projection | |
| CN109040497B (en) | Proportional affine projection self-adaptive echo cancellation method based on M estimation | |
| CN104601837A (en) | Robust convex combination type adaptive phone echo canceling method | |
| CN107749303B (en) | Post-processing method and device for voice signal output by acoustic echo canceller | |
| CN105491256B (en) | A kind of acoustic echo canceller startup stage steady step length regulating method | |
| CN110191245B (en) | Self-adaptive echo cancellation method based on time-varying parameters | |
| CN105721729A (en) | Echo cancellation method based on block-sparse proportionate reuse weight coefficient affine projection | |
| CN104144269A (en) | Proportional self-adaption telephone echo cancellation method based on decorrelation | |
| CN105070295A (en) | Adaptive method, applied to echo cancellation, of active factor proportional sub band | |
| CN107871510A (en) | A kind of zero attracts to become the maximum entropy echo cancel method of core width | |
| CN114283830A (en) | Deep learning network-based microphone signal echo cancellation model construction method | |
| CN106170049B (en) | A kind of normalization sub-band adaptive echo cancel method having offset compensation | |
| Shen et al. | An improved LMS adaptive filtering algorithm and its analysis | |
| CN106157965A (en) | A kind of zero norm collection person's illumination-imitation projection self-adoptive echo cancel method reused based on weight vector | |
| CN106161822B (en) | A kind of normalization sub-band adaptive echo cancel method based on M estimation | |
| CN111739548B (en) | Minimum error entropy self-adaptive echo cancellation method for resisting input noise | |
| JP2002503910A (en) | Control method and apparatus for filter adaptation in noise environment | |
| CN106060295A (en) | Proportional affine projection echocancellation method of convex combination coefficient differences |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20170215 Assignee: Yangjiang taiyijiehua Technology Co.,Ltd. Assignor: DONGGUAN University OF TECHNOLOGY Contract record no.: X2023990000090 Denomination of invention: A time-frequency memory subband proportional adaptive echo cancellation method Granted publication date: 20190614 License type: Common License Record date: 20230112 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20170215 Assignee: Dongguan Zongyou Technology Information Service Co.,Ltd. Assignor: DONGGUAN University OF TECHNOLOGY Contract record no.: X2023980035041 Denomination of invention: A Time Frequency Memory Subband Proportional Adaptive Echo Cancellation Method Granted publication date: 20190614 License type: Common License Record date: 20230424 |
|
| EE01 | Entry into force of recordation of patent licensing contract |