CN111050005B - Bias compensation collective affine projection echo cancellation method - Google Patents
Bias compensation collective affine projection echo cancellation method Download PDFInfo
- Publication number
- CN111050005B CN111050005B CN201911317433.9A CN201911317433A CN111050005B CN 111050005 B CN111050005 B CN 111050005B CN 201911317433 A CN201911317433 A CN 201911317433A CN 111050005 B CN111050005 B CN 111050005B
- Authority
- CN
- China
- Prior art keywords
- vector
- signal
- adaptive filter
- current time
- echo
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
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
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
The invention relates to an echo cancellation technology, belonging to the field of telephone communication. The method discloses an echo cancellation method of offset compensation set element radiation projection, and solves the problems that the algorithm complexity is high and the echo cancellation has offset when the input signal is mixed with noise in the traditional echo cancellation algorithm. The invention calculates the deviation compensation vector of the current moment and takes the compensation vector as a reduction term when the weight coefficient vector is updated. When the weight coefficient vector is larger, the subtraction term value is also larger, so that the faster initial convergence speed is obtained; the subtraction value is reduced when the tap weight coefficient is close to the steady state, so that the updating speed of the tap weight coefficient is correspondingly reduced when the tap weight coefficient is close to the steady state, and better stability is kept.
Description
Technical Field
The invention relates to an echo cancellation technology, in particular to an echo cancellation method for deviation compensation collective member radiation projection, and belongs to the field of telephone communication.
Background
In a real-time communication system, real-time voice communication between two or more parties is required. Echo during real-time voice communication is a big problem of interfering with communication, for example, if the voice of the user is heard from the receiver during telephone communication, the echo is generated during communication.
At present, with the opening of various multifunctional communication services, echo problems inevitably occur in all occasions requiring the simultaneous use of a loudspeaker and a microphone, such as a hands-free telephone system, an IP telephone, a video teleconference system and a voice control system. The human ear is extremely sensitive to the echo, the echo delayed for 10ms can be captured and sensed by the human ear in the communication process, and the echo exceeding 32ms causes great interference to the communication. If the echo signal is not processed, the call quality will be affected, and oscillation will be formed seriously, resulting in howling. Therefore, effective measures must be taken to suppress the echo signal and improve the voice call quality.
Echo cancellation requires estimation of the characteristic parameters of the echo path, and since the echo path is usually unknown and time-varying, there are some difficulties in estimating the echo path, so at present, an adaptive algorithm is generally adopted to simulate the echo path. When the statistical properties of the input signal change, the adaptive algorithm can track the change and automatically adjust the parameters to optimize the filter performance again. Compared with other echo cancellation methods, the self-adaptive echo cancellation technology has the characteristics of low application cost, high convergence speed and small echo residual error. It is the most promising echo cancellation technology internationally acknowledged at present, and is also the mainstream technology adopted by echo cancellation at present.
The basic principle of the adaptive echo cancellation technology is based on the correlation between the loudspeaker signal and the multipath echo generated by the loudspeaker signal, a speech model of the far-end signal is established, the echo is estimated by using the speech model, and the coefficient of a filter is continuously modified, so that the estimated value is closer to the real echo. The echo estimate is then subtracted from the input signal of the microphone to cancel the echo. Therefore, how to perfect and research the adaptive echo cancellation algorithm with excellent performance is the research core in the field of echo cancellation. In the current application of sparse system identification, the following two methods are more mature:
(1) the method can be regarded as an extension of a Normalized Least Mean Square (NLMS) algorithm and has a high convergence rate. Affine Projection (AP) algorithm is one of the classical adaptive echo cancellation algorithms that are used more frequently at present, however, the algorithm complexity is high.
(2) In order to reduce the computational complexity of an AP algorithm, a learner provides an ensemble affine projection (SM-AP) algorithm based on the ensemble affine projection (SM-AP) algorithm. However, this method does not consider the case where the input signal contains noise, and when the input signal contains noise, the SM-AP algorithm is biased in echo cancellation.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the echo cancellation method for the offset compensation centralized member radiation projection is provided, and the problems that the algorithm complexity is high and the echo cancellation has offset when the input signals are mixed with noise in the traditional echo cancellation algorithm are solved.
The technical scheme adopted by the invention for solving the technical problems is as follows:
an echo cancellation method for bias compensated affine projection of a member, comprising the steps of:
A. a far-end signal sampling step:
sampling the far-end signal transmitted from the far-end to obtain the far-end signal discrete value x (n) of the current time n,
and (3) forming the remote signal discrete values x (n), x (n-1) of the current time n and L-1 previous times into an adaptive filter input vector x' (n) of the current time n:
x′(n)＝[x(n),x(n-1),...,x(n-L+1)]Tl is the tap length of the adaptive filter;
forming an input matrix X (n) with L multiplied by P dimensions by using input vectors at n, n-1, …, n-P +1 time:
x (n) ═ X ' (n), X ' (n-1), …. X ' (n-P +1) ], P being the order of the projection of the radiation;
B. echo signal estimation step:
superposing the noise vector eta (n) of the current time n with the input vector x' (n) of the adaptive filter to form the input vector of the adaptive filter containing noise
η(n)＝[η(n),η(n-1),...,η(n-L+1)]T；
The noisy input matrix of dimension L x P is
The noisy input moment of L multiplied by P dimensionMatrix of
Obtaining a P multiplied by 1 dimensional output vector y (n) of the current time n, namely an estimation vector y (n) of the echo signal through an adaptive filter:
wherein w (n) ═ w1(n),w2(n),...,wl(n),...,wL(n)]TThe adaptive filter tap weight vector for the current time n has an initial value of zero, wl(n) is the l tap weight coefficient of the adaptive filter at the current time n;
C. echo signal eliminating step:
sampling the near-end signal to obtain a near-end sampling signal d (n) with echo at the current moment n;
subtracting the echo near-end sampling signal d (n) at the current moment n from the estimated value y (n) of the echo signal at the current moment n of the adaptive filter to obtain a residual signal e (n) at the current moment n:
e(n)＝d(n)-y(n)＝[e0(n),e1(n),…,eP-1(n)]T；
then the residual error signal e (n) of the current moment n is sent back to the far end;
D. the filter tap weight coefficient updating step comprises the steps D1-D2:
d1, calculating a deviation compensation vector:
calculating a deviation compensation vector rho (n) of the current time n from a residual signal e (n) of the current time n and an adaptive filter tap weight coefficient vector w (n) of the current time n:
where gamma is the residual e0(n) an upper bound constraint value, Tr (-) represents a trace of the matrix,
the variance of the noise vector η (n) for the current time instant n;
d2, update of tap weight coefficient vector:
the adaptive filter tap weight coefficient vector w (n +1) at the next time instant n +1 is updated as follows:
E. and c, enabling n to be n +1, and repeatedly executing the steps A-D until the call is ended.
As a further optimization, in step a, the tap length L of the adaptive filter is 128.
As a further optimization, in step a, the value of the radiation projection order P is 4.
The invention has the beneficial effects that:
in the SM-AP algorithm, for noise signals mixed in input signals, a deviation compensation vector at the current moment is calculated through a residual signal at the current moment and a tap weight coefficient vector of an adaptive filter at the current moment, and the compensation vector is used as a reduction term when the weight coefficient vector is updated. When the weight coefficient vector is larger, the subtraction term value is also larger, so that the faster initial convergence speed is obtained; the subtraction value is reduced when the tap weight coefficient is close to the steady state, so that the updating speed of the tap weight coefficient is correspondingly reduced when the tap weight coefficient is close to the steady state, and better stability is kept.
Drawings
FIG. 1 is a flow chart of the steps of the echo cancellation method of the present invention;
FIG. 2 is a normalized steady-state imbalance curve using the AP method, SM-AP method, and the method, respectively, in a simulation experiment.
Detailed Description
The invention aims to provide a bias-compensated echo cancellation method for collective member radiation projection, which solves the problems that the algorithm complexity is higher and the echo cancellation has bias when the input signals are mixed with noise in the traditional echo cancellation algorithm. The core idea is as follows: and calculating a deviation compensation vector at the current moment, and taking the compensation vector as a reduction term when the weight coefficient vector is updated. When the weight coefficient vector is larger, the subtraction term value is also larger, so that the faster initial convergence speed is obtained; the subtraction value is reduced when the tap weight coefficient is close to the steady state, so that the updating speed of the tap weight coefficient is correspondingly reduced when the tap weight coefficient is close to the steady state, and better stability is kept.
In a specific implementation, as shown in fig. 1, the echo cancellation method for deviation-compensated collective projection in the present invention includes the following steps:
A. a far-end signal sampling step:
sampling the far-end signal transmitted from the far-end to obtain the far-end signal discrete value x (n) of the current time n,
and (3) forming the remote signal discrete values x (n), x (n-1) of the current time n and L-1 previous times into an adaptive filter input vector x' (n) of the current time n:
x′(n)＝[x(n),x(n-1),...,x(n-L+1)]Tl is the tap length of the adaptive filter, taking value 128;
forming an input matrix X (n) with L multiplied by P dimensions by using input vectors at n, n-1, …, n-P +1 time:
x (n) ([ X ' (n), X ' (n-1), …. X ' (n-P +1) ], P being the projection order, value 4;
B. echo signal estimation step:
superposing the noise vector eta (n) of the current time n with the input vector x' (n) of the adaptive filter to form the input vector of the adaptive filter containing noise
η(n)＝[η(n),η(n-1),...,η(n-L+1)]T；
The noisy input matrix of dimension L x P is
Inputting the noisy input matrix with L multiplied by P dimension
Obtaining a P multiplied by 1 dimensional output vector y (n) of the current time n, namely an estimation vector y (n) of the echo signal through an adaptive filter:
wherein w (n) ═ w1(n),w2(n),...,wl(n),...,wL(n)]TThe adaptive filter tap weight vector for the current time n has an initial value of zero, wl(n) is the l tap weight coefficient of the adaptive filter at the current time n;
C. echo signal eliminating step:
sampling the near-end signal to obtain a near-end sampling signal d (n) with echo at the current moment n;
subtracting the echo near-end sampling signal d (n) at the current moment n from the estimated value y (n) of the echo signal at the current moment n of the adaptive filter to obtain a residual signal e (n) at the current moment n:
e(n)＝d(n)-y(n)＝[e0(n),e1(n),…,eP-1(n)]T；
then the residual error signal e (n) of the current moment n is sent back to the far end;
D. the filter tap weight coefficient updating step comprises the steps D1-D2:
d1, calculating a deviation compensation vector:
calculating a deviation compensation vector rho (n) of the current time n from a residual signal e (n) of the current time n and an adaptive filter tap weight coefficient vector w (n) of the current time n:
where gamma is the residual e0(n) an upper bound constraint value, Tr (-) represents a trace of the matrix,
the variance of the noise vector η (n) for the current time instant n;
d2, update of tap weight coefficient vector:
the adaptive filter tap weight coefficient vector w (n +1) at the next time instant n +1 is updated as follows:
E. and c, enabling n to be n +1, and repeatedly executing the steps A-D until the call is ended.
Simulation experiment:
in order to verify the effectiveness of the method, the performance of the echo cancellation method is compared with the traditional AP algorithm and SM-AP algorithm through experiments.
In the simulation experiment, the tap length L of the adaptive filter is 128, a first-order autoregressive (AR (1)) signal is adopted as a far-end input signal, and after the received far-end signal is played by a loudspeaker in a quiet closed room with the room length of 6.25m, the width of 3.75m, the height of 2.5m, the temperature of 20 ℃ and the humidity of 50 percent, the received far-end signal is collected by a microphone in the room according to the sampling frequency of 8000 Hz.
The parameters of each algorithm in the experiment are specifically as follows:
parameter value of each method experiment
| AP method | μ＝0.2 |
| SM-AP method | γ＝0.05 |
| The invention | γ＝0.05 |
The simulation results were obtained by running 100 averages independently.
It can be seen from fig. 2 that in a sparse system environment, the steady state detuning of the present invention is about-22 dB, significantly lower than-16 dB for the AP method and-18 dB for the SM-AP method, at the same convergence rate. Therefore, the scheme of the invention is obviously superior to the traditional technology in the stability of noise elimination.
Claims (3)
1. An echo cancellation method for bias compensated affine projection of a member, comprising the steps of:
A. a far-end signal sampling step:
sampling the far-end signal transmitted from the far-end to obtain the far-end signal discrete value x (n) of the current time n,
and (3) forming the remote signal discrete values x (n), x (n-1) of the current time n and L-1 previous times into an adaptive filter input vector x' (n) of the current time n:
x′(n)＝[x(n),x(n-1),...,x(n-L+1)]Tl is the tap length of the adaptive filter;
forming an input matrix X (n) with L multiplied by P dimensions by using input vectors at n, n-1, …, n-P +1 time:
x (n) ═ X ' (n), X ' (n-1),.. X ' (n-P +1) ], P being the radiation projection order;
B. echo signal estimation step:
superposing the noise vector eta (n) of the current time n with the input vector x' (n) of the adaptive filter to form the input vector of the adaptive filter containing noise
η(n)＝[η(n),η(n-1),...,η(n-L+1)]T；
The noisy input matrix of dimension L x P is
Inputting the noisy input matrix with L multiplied by P dimension
Obtaining a P multiplied by 1 dimensional output vector y (n) of the current time n, namely an estimation vector y (n) of the echo signal through an adaptive filter:
wherein w (n) ═ w1(n),w2(n),...,wl(n),...,wL(n)]TThe adaptive filter tap weight vector for the current time n has an initial value of zero, wl(n) is the l tap weight coefficient of the adaptive filter at the current time n;
C. echo signal eliminating step:
sampling the near-end signal to obtain a near-end sampling signal d (n) with echo at the current moment n;
subtracting the echo near-end sampling signal d (n) at the current moment n from the estimated value y (n) of the echo signal at the current moment n of the adaptive filter to obtain a residual signal e (n) at the current moment n:
e(n)＝d(n)-y(n)＝[e0(n),e1(n),…,eP-1(n)]T；
then the residual error signal e (n) of the current moment n is sent back to the far end;
D. the filter tap weight coefficient updating step comprises the steps D1-D2:
d1, calculating a deviation compensation vector:
calculating a deviation compensation vector rho (n) of the current time n from a residual signal e (n) of the current time n and an adaptive filter tap weight coefficient vector w (n) of the current time n:
where gamma is the residual e0(n) an upper bound constraint value, Tr (-) represents a trace of the matrix,
the variance of the noise vector η (n) for the current time instant n;
d2, update of tap weight coefficient vector:
the adaptive filter tap weight coefficient vector w (n +1) at the next time instant n +1 is updated as follows:
E. and c, enabling n to be n +1, and repeatedly executing the steps A-D until the call is ended.
2. The method of claim 1, wherein said offset-compensated affine projection echo cancellation method,
in step a, the tap length L of the adaptive filter is 128.
3. The method of claim 1, wherein said offset-compensated affine projection echo cancellation method,
in the step A, the value of the radiation projection order P is 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911317433.9A CN111050005B (en) | 2019-12-19 | 2019-12-19 | Bias compensation collective affine projection echo cancellation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911317433.9A CN111050005B (en) | 2019-12-19 | 2019-12-19 | Bias compensation collective affine projection echo cancellation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111050005A CN111050005A (en) | 2020-04-21 |
| CN111050005B true CN111050005B (en) | 2021-02-09 |
Family
ID=70237795
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911317433.9A Active CN111050005B (en) | 2019-12-19 | 2019-12-19 | Bias compensation collective affine projection echo cancellation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111050005B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107610714A (en) * | 2017-09-13 | 2018-01-19 | 西南交通大学 | The echo cancel method of the minimum cube absolute value attracted based on a norm zero |
| 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 |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6909782B2 (en) * | 2000-09-08 | 2005-06-21 | Intel Corporation | Fast converging affine projection based echo canceller for sparse multi-path channels |
| CN103561185B (en) * | 2013-11-12 | 2015-08-12 | 沈阳工业大学 | A kind of echo cancel method of sparse path |
| CN104683614B (en) * | 2015-03-24 | 2016-03-02 | 西南交通大学 | Based on the proportional illumination-imitation projection self-adoptive echo cancel method of memory that M estimates |
| CN109697986B (en) * | 2018-09-19 | 2020-12-18 | 四川大学 | Adaptive bias compensation echo cancellation method based on minimum cubic absolute value |
-
2019
- 2019-12-19 CN CN201911317433.9A patent/CN111050005B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107610714A (en) * | 2017-09-13 | 2018-01-19 | 西南交通大学 | The echo cancel method of the minimum cube absolute value attracted based on a norm zero |
| 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 |
Non-Patent Citations (2)
| Title |
|---|
| A Bias-Compensated Affine Projection Algorithm;Sang Mok Jung，Nam Kyu Kwon，PooGyeon Park;《Control conference（ASCC）2013 9th Asia》;20130630;全文 * |
| 基于集员滤波的变步长仿射投影算法;张晓丽，范永全，张家树;《西华大学学报》;20110731;第30卷(第4期);80-82 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111050005A (en) | 2020-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105825864B (en) | Both-end based on zero-crossing rate index is spoken detection and echo cancel method | |
| CN109754813B (en) | Variable step size echo cancellation method based on rapid convergence characteristic | |
| CN102461205B (en) | Acoustic multi-channel echo cancellation device and method for cancelling acoustic multi-channel echo | |
| JP5284475B2 (en) | Method for determining updated filter coefficients of an adaptive filter adapted by an LMS algorithm with pre-whitening | |
| Gil-Cacho et al. | Nonlinear acoustic echo cancellation based on a sliding-window leaky kernel affine projection algorithm | |
| CN106533500B (en) | A method of optimization Echo Canceller convergence property | |
| US8934620B2 (en) | Acoustic echo cancellation for high noise and excessive double talk | |
| Costa et al. | Acoustic echo cancellation using nonlinear cascade filters | |
| WO2020181766A1 (en) | Voice signal processing method and device, apparatus, and readable storage medium | |
| CN109769060A (en) | A kind of mobile phone active noise reducing device and method | |
| CN107134281A (en) | Adaptive filter coefficient update method during a kind of adaptive echo is eliminated | |
| CN110191245B (en) | Self-adaptive echo cancellation method based on time-varying parameters | |
| CN109102794A (en) | M based on convex combination estimates the echo cancel method of proportional class affine projection | |
| Gil-Cacho et al. | Wiener variable step size and gradient spectral variance smoothing for double-talk-robust acoustic echo cancellation and acoustic feedback cancellation | |
| CN109697986B (en) | Adaptive bias compensation echo cancellation method based on minimum cubic absolute value | |
| CN109040497B (en) | Proportional affine projection self-adaptive echo cancellation method based on M estimation | |
| CN107071196B (en) | A kind of adaptive echo cancellation method | |
| CN113873090B (en) | Robust estimation affine projection spline self-adaptive echo cancellation method | |
| JP2003324372A (en) | Improved acoustic echo cancellation | |
| CN111050005B (en) | Bias compensation collective affine projection echo cancellation method | |
| CN113409806B (en) | Zero-attraction echo cancellation method based on arctangent function | |
| CN110767245B (en) | Voice communication self-adaptive echo cancellation method based on S-shaped function | |
| Song et al. | A nonparametric variable step-size subband adaptive filtering algorithm for acoustic echo cancellation | |
| CN109040498B (en) | Method and system for improving echo cancellation effect | |
| Raghavendran | Implementation of an acoustic echo canceller using matlab |
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 |