[go: up one dir, main page]

CN108877824B - A Combined Step-Size Echo Cancellation Method with High Tracking Performance - Google Patents

A Combined Step-Size Echo Cancellation Method with High Tracking Performance Download PDF

Info

Publication number
CN108877824B
CN108877824B CN201810547634.7A CN201810547634A CN108877824B CN 108877824 B CN108877824 B CN 108877824B CN 201810547634 A CN201810547634 A CN 201810547634A CN 108877824 B CN108877824 B CN 108877824B
Authority
CN
China
Prior art keywords
signal
error
adaptive filter
current time
value
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
Application number
CN201810547634.7A
Other languages
Chinese (zh)
Other versions
CN108877824A (en
Inventor
赵海全
施龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Hongyue Information Technology Co ltd
Original Assignee
Southwest Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Southwest Jiaotong University filed Critical Southwest Jiaotong University
Priority to CN201810547634.7A priority Critical patent/CN108877824B/en
Publication of CN108877824A publication Critical patent/CN108877824A/en
Application granted granted Critical
Publication of CN108877824B publication Critical patent/CN108877824B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise
    • 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)
  • Computational Linguistics (AREA)
  • Quality & Reliability (AREA)
  • Signal Processing (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

一种跟踪性能高的组合步长回声消除方法，其步骤包括：A、采集远端输入信号u(n)和近端期望信号d(n)；B、将输入信号离散值u(n)输入自适应滤波器，得到输出信号y(n)；C、用d(n)减去y(n)进行回声消除，将误差信号e(n)传送给远端；D、计算自适应滤波器的误差向量；E、更新自适应滤波器的抽头权值w(n)；如当前时刻的误差信号的相对强度值

Figure DDA0001680195310000011
大于设定的阈值，且前一时刻的误差信号的绝对强度值r(n‑1)＝log10(e2(n‑1))小于设定的阈值，则将权值误差矩阵迹J(n)强制重置为一个大的初始值；F、更新权值误差矩阵迹J(n+1)；G、令n＝n+1，重复步骤A、B、C、D、E、F，直至通话结束。该方法在声学回声信道发生变化时，具有良好的跟踪性能。

Figure 201810547634

A combined step size echo cancellation method with high tracking performance, the steps of which include: A. collecting a far-end input signal u(n) and a near-end desired signal d(n); B. inputting the discrete value u(n) of the input signal Adaptive filter to obtain the output signal y(n); C. Use d(n) to subtract y(n) for echo cancellation, and transmit the error signal e(n) to the far end; D. Calculate the value of the adaptive filter Error vector; E, update the tap weight w(n) of the adaptive filter; such as the relative strength value of the error signal at the current moment

Figure DDA0001680195310000011
is greater than the set threshold, and the absolute strength value of the error signal at the previous moment r(n-1)=log 10 (e 2 (n-1)) is less than the set threshold, then the weight error matrix trace J( n) Force reset to a large initial value; F, update weight error matrix trace J(n+1); G, set n=n+1, repeat steps A, B, C, D, E, F, until the call ends. The method has good tracking performance when the acoustic echo channel changes.

Figure 201810547634

Description

Combined step echo cancellation method with high tracking performance
Technical Field
The invention belongs to the technical field of echo cancellation in communication.
Background
Adaptive signal processing is an important branch of information technology and is widely used in the field of communications. In the field of communications, echo cancellation is a very interesting and challenging hotspot. Multiple reflections of sound in an enclosed space can create echoes, as well as echoes in the signal transmission due to impedance mismatches in the transmission medium. Communication echoes can be cancelled by a system identification model: the recognized system is an acoustic echo channel, the output of the system recognition is the estimation of an echo signal, and the echo cancellation can be realized by subtracting the estimation of the echo signal from the voice signal containing the echo signal, which is the principle of the adaptive echo canceller.
In echo cancellation applications, the speech signal has a correlation characteristic. In the case of correlated input signals, the performance of conventional adaptive filtering algorithms such as Least Mean Square (LMS) and Normalized Least Mean Square (NLMS) is significantly degraded. For this purpose, the affine projection algorithm (h.c. shin, a.h. sayed, Mean-square performance of a family of fine projection algorithms [ J ], IEEE trans. signal process, 2004,52(1), pp.90-102), abbreviated as APA, is proposed by h.c. shin. The affine projection algorithm updates the weight by using an input matrix formed by a plurality of input vectors, and has the capability of decorrelating signals, thereby accelerating the convergence speed of the algorithm. However, the fixed-step APA algorithm suffers from a tradeoff of fast convergence speed and low steady-state imbalance. To solve this conflict, j.h.choi proposes a scheme of combining step sizes (j.h.choi, h.cho, s.w.kim, Combination of step sizes for after-field project algorithm with variable missing parameter, electron.let, 2013,49(18), pp.1149-1150), abbreviated as CSSAPA-VMP. The CSSAPA-VMP algorithm combines a large step size and a small step size by using an adaptive mixing parameter to ensure that the algorithm can obtain a fast convergence speed in an initial stage and obtain low steady state maladjustment in a steady state stage.
The tracking capability is an important index for evaluating the quality of the adaptive filtering algorithm, and has practical significance for echo cancellation application. However, the mixing parameters of the CSSAPA-VMP algorithm mainly depend on the input vector or the variance of the echo signals formed by the input signals at the current moment and the previous moments, the input signal proportion of the current moment in the input vector is lower, and the echo signal proportion of the current moment in the variance is also lower; therefore, when the channel changes suddenly, it has no tracking capability and the performance becomes very poor, resulting in undesirable echo cancellation effect.
Disclosure of Invention
The invention aims to provide a combined step size echo cancellation method with high tracking performance, which has good tracking performance and can still obtain high convergence speed and low steady state maladjustment when an acoustic echo channel changes.
The technical scheme adopted by the invention for realizing the aim is that the combined step size echo cancellation method with high tracking performance comprises the following steps:
A. signal acquisition
Sampling a far-end signal transmitted from a far end to obtain an input signal discrete value u (n) of the current moment n, and sampling a near-end signal to obtain an expected signal discrete value d (n) of the current moment n with echo;
B. echo signal estimation
The discrete value u (n) of the input signal at the time n to n-L +1, u (n-1).. u (n-L +1), and the adaptive filter input vector U (n) at the current time n, U (n) [ (n), u (n-1).. u (n-L +1) ], u (n-L +1)]TWhere L denotes the length of the adaptive filter, L is 512 or 1024, and the superscript T denotes transposition;
inputting the discrete value u (n) of the input signal into the adaptive filter to obtain the output signal y (n) of the adaptive filter at the current time n, wherein y (n) is WT(n) U (n); wherein, W (n) is the weight vector of the adaptive filter at the current moment n, the length of the weight vector is equal to L, and the initial value is a zero vector;
C. echo cancellation
Subtracting the output signal y (n) of the adaptive filter at the current time n from the desired signal discrete value d (n) at the current time n obtained in the step a to obtain an error signal e (n) at the current time n, namely e (n) ═ d (n) -y (n); and transmitting the error signal e (n) of the current moment n to the far end as a pure signal after echo cancellation;
D. adaptive filter error vector calculation
D1, using the input vector U (n), U (n-1) and U (n-P +1) of the adaptive filter from the current time n to the time n-P +1 to form the input matrix of the adaptive filter at the current time n
Figure BDA0001680195290000031
Figure BDA0001680195290000032
Wherein P is an affine projection order, and the value range of P is 2-6;
d2, using the desired signal D (n), D (n-1),. and D (n-P +1) from the current time n to the time n-P +1 to form the desired signal vector D (n) at the current time n, D (n) ([ D (n), D (n-1),. and D (n-P +1)]T；
D3, subtracting the input matrix of the current time n from the expected signal vector D (n) of the current time n
Figure BDA0001680195290000033
The transpose of (a) and the product of the weight vector W (n) of the adaptive filter to obtain the error vector E (n) of the adaptive filter at the current time n,
Figure BDA0001680195290000034
E. adaptive filter tap weight update
E1, calculating a weight error vector v (n) of the current time n, where v (n) is W ° -W (n), where W ° is an expected value of the weight vector of the adaptive filter, that is, an echo channel to be estimated; calculating weight error matrix Q (n) at current time n, Q (n) V (n)T；
E2, calculation of mixing parameters
Calculating weight error matrix trace J (n) at current time n, wherein J (n) is Tr (Q (n)), and Tr (·) represents trace of matrix calculation
Using the filter length L in step B, the input matrix in step D1
Figure BDA0001680195290000035
And affine projection order P, calculating a mixing parameter λ (n) of the current time n:
Figure BDA0001680195290000041
wherein, mu2Is a small step size, mu1The step length is large, the value ranges of the two are both 0.0001 to 0.5,
Figure BDA0001680195290000042
the variance of the echo signal is within the range of 0.001-0.1;
e3, triggering of reset mechanism
Calculating the relative strength of the signal error at the current time n
Figure BDA0001680195290000043
Simultaneously calculating the absolute intensity r (n-1) of the signal error of the previous time n-1 as log10(e2(n-1))；
If R (n) ≦ t1Or R (n) > t1And r (n-1) is not less than t2If so, judging that the echo channel does not have mutation at the current moment, and performing the operation of step E4; wherein, t1The threshold value of the relative strength of the signal error is represented, the value range is 1-3, t2A threshold value representing the absolute intensity of the signal error, the value range is-3 to-1, q0To representResetting the initial value, wherein the value range is 1-10;
if R (n) > t1While r (n-1) < t2If the echo channel is suddenly changed, the weight error matrix trace is reset, i.e. J (n) q0；
E4 updating of filter tap weights
Using the input matrix in step D1
Figure BDA0001680195290000044
The error vector E (n) in step D3 and the blending parameter λ (n) in step E2 obtain the filter tap weight W (n +1) at the next time (n + 1):
Figure BDA0001680195290000045
wherein, (.)-1Representing an inversion matrix;
F. updating weight error matrix trace
Using the filter length L in step B, the input matrix in step D1
Figure BDA0001680195290000046
And affine projection order P, mixing the parameter λ (n) and the weight error matrix q (n) in step E2, and obtaining a weight error matrix trace J (n +1) at the next time (n + 1):
Figure BDA0001680195290000051
G. repetition of
Let n be n +1, repeat step A, B, C, D, E, F until the call ends.
Compared with the prior art, the invention has the beneficial effects that:
the invention designs a reset mechanism aiming at the change of an acoustic echo channel, which is used for detecting whether the channel has sudden change at each moment. In particular if the relative strength value of the error signal at the current time is
Figure BDA0001680195290000052
Greater than a set threshold value, and the absolute intensity value r (n-1) of the error signal at the previous moment is log10(e2(n-1)) is smaller than a set threshold, the echo channel at the current moment is judged to generate sudden change (the channel at the previous moment is normal); then, the weight error matrix trace J (n) is forcibly reset to a large initial value, further, the mixing parameter is immediately reset to a large numerical value, and a large-step filter in the combined filter plays a leading role, so that the algorithm of the invention has strong tracking capability when the echo channel is suddenly changed, and can obtain high convergence speed and low steady-state imbalance.
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
Drawings
FIG. 1 is a speech signal of a simulation experiment of the present invention;
FIG. 2 is a normalized misadjustment curve of the present invention under speech input.
Detailed Description
Examples
A specific embodiment of the present invention is a combined step echo cancellation method with high tracking performance, which includes the following steps:
A. signal acquisition
Sampling a far-end signal transmitted from a far end to obtain an input signal discrete value u (n) of the current moment n, and sampling a near-end signal to obtain an expected signal discrete value d (n) of the current moment n with echo;
B. echo signal estimation
The discrete value u (n) of the input signal at the time n to n-L +1, u (n-1).. u (n-L +1), and the adaptive filter input vector U (n) at the current time n, U (n) [ (n), u (n-1).. u (n-L +1) ], u (n-L +1)]TWhere L denotes the length of the adaptive filter, L is 512 or 1024, and the superscript T denotes transposition;
inputting the discrete value u (n) of the input signal into the adaptive filter to obtain the output signal y (n) of the adaptive filter at the current time n, wherein y (n) is WT(n) U (n); wherein W: (n) is a weight vector of the adaptive filter at the current moment n, the length of the weight vector is equal to L, and the initial value is a zero vector;
C. echo cancellation
Subtracting the output signal y (n) of the adaptive filter at the current time n from the desired signal discrete value d (n) at the current time n obtained in the step a to obtain an error signal e (n) at the current time n, namely e (n) ═ d (n) -y (n); and transmitting the error signal e (n) of the current moment n to the far end as a pure signal after echo cancellation;
D. adaptive filter error vector calculation
D1, using the input vector U (n), U (n-1) and U (n-P +1) of the adaptive filter from the current time n to the time n-P +1 to form the input matrix of the adaptive filter at the current time n
Figure BDA0001680195290000061
Figure BDA0001680195290000062
Wherein P is an affine projection order, and the value range of P is 2-6;
d2, using the desired signal D (n), D (n-1),. and D (n-P +1) from the current time n to the time n-P +1 to form the desired signal vector D (n) at the current time n, D (n) ([ D (n), D (n-1),. and D (n-P +1)]T；
D3, subtracting the input matrix of the current time n from the expected signal vector D (n) of the current time n
Figure BDA0001680195290000063
The transpose of (a) and the product of the weight vector W (n) of the adaptive filter to obtain the error vector E (n) of the adaptive filter at the current time n,
Figure BDA0001680195290000064
E. adaptive filter tap weight update
E1, calculating a weight error vector v (n) of the current time n, where v (n) is W ° -W (n), where W ° is an expected value of the weight vector of the adaptive filter, that is, an echo channel to be estimated; then, the product is processedCalculating a weight error matrix Q (n) at the current time n, wherein Q (n) is V (n)T；
E2, calculation of mixing parameters
Calculating weight error matrix trace J (n) at current time n, wherein J (n) is Tr (Q (n)), and Tr (·) represents trace of matrix calculation
Using the filter length L in step B, the input matrix in step D1
Figure BDA0001680195290000071
And affine projection order P, calculating a mixing parameter λ (n) of the current time n:
Figure BDA0001680195290000072
wherein, mu2Is a small step size, mu1The step length is large, the value ranges of the two are both 0.0001 to 0.5,
Figure BDA0001680195290000073
the variance of the echo signal is within the range of 0.001-0.1;
e3, triggering of reset mechanism
Calculating the relative strength of the signal error at the current time n
Figure BDA0001680195290000074
Simultaneously calculating the absolute intensity r (n-1) of the signal error of the previous time n-1 as log10(e2(n-1))；
If R (n) ≦ t1Or R (n) > t1And r (n-1) is not less than t2If so, judging that the echo channel does not have mutation at the current moment, and performing the operation of step E4; wherein, t1The threshold value of the relative strength of the signal error is represented, the value range is 1-3, t2A threshold value representing the absolute intensity of the signal error, the value range is-3 to-1, q0Representing a reset initial value, wherein the value range is 1-10;
if R (n) > t1While r (n-1) < t2If so, the echo channel at the current moment is judged to have sudden change, and the weight value is adjustedResetting the error matrix trace, i.e. making J (n) q0；
E4 updating of filter tap weights
Using the input matrix in step D1
Figure BDA0001680195290000081
The error vector E (n) in step D3 and the blending parameter λ (n) in step E2 obtain the filter tap weight W (n +1) at the next time (n + 1):
Figure BDA0001680195290000082
wherein, (.)-1Representing an inversion matrix;
F. updating weight error matrix trace
Using the filter length L in step B, the input matrix in step D1
Figure BDA0001680195290000083
And affine projection order P, mixing the parameter λ (n) and the weight error matrix q (n) in step E2, and obtaining a weight error matrix trace J (n +1) at the next time (n + 1):
Figure BDA0001680195290000084
G. repetition of
Let n be n +1, repeat step A, B, C, D, E, F until the call ends.
Simulation experiment
To verify the effectiveness of the present invention, we performed simulation experiments.
In the simulation experiment, the impulse response of the echo channel is formed in a quiet closed room with the length M of 512, wherein the room is 6.25M long, 3.75M wide and 2.5M high, the temperature is 20 ℃ and the humidity is 50%. The real voice signal played by the loudspeaker is adopted as the input signal of the far end, the sampling frequency is 8000Hz, and the number of sampling points is 80000, as shown in figure 1.
Echo signals are collected from a telephone microphone, and the impulse response of an echo channel changes abruptly at a position of half the number of sampling points.
In the simulation experiment, the projection order value of all tested algorithms is P-4, and the step length value is marked in the simulation graph. For the CSSAPA-CMP algorithm,
Figure BDA0001680195290000085
with respect to the present invention, it is,
Figure BDA0001680195290000086
q0＝5，t1＝2，t2＝-2。
FIG. 2 is a simulation of the APA algorithm, CSSAPA-VMP algorithm and the normalized misadjustment of the present invention under a speech input signal. As can be seen from fig. 2, before the echo channel suddenly changes (before the sampling time 40000), the performance of the present invention is close to that of the CSSAPA-VMP algorithm, and both the performance of the present invention and the performance of the rsa algorithm are superior to that of the APA algorithm. After the echo channel has sudden change (sampling time 40000), the performance of the APA algorithm and the CSSAPA-VMP algorithm is poor, and the weight estimated by the algorithm deviates far from the echo channel and cannot track the sudden change of the echo channel. In contrast, when the echo channel is mutated, the weight estimation value of the algorithm is correspondingly changed, is very close to the echo channel and has strong tracking capability; the convergence rate is high and the steady state imbalance is low.

Claims (1)

1.一种跟踪性能高的组合步长回声消除方法，其步骤如下：1. A combined step size echo cancellation method with high tracking performance, the steps of which are as follows: A、信号采集A. Signal acquisition 将远端传来的远端信号采样得到当前时刻n的输入信号离散值u(n)，对近端信号采样得到带有回声的当前时刻n的期望信号离散值d(n)；Sampling the far-end signal from the far end to obtain the discrete value u(n) of the input signal at the current time n, and sampling the near-end signal to obtain the desired signal discrete value d(n) at the current time n with echo; B、回声信号估计B. Echo signal estimation 将输入信号离散值u(n)在n到n-L+1时刻的值u(n),u(n-1)...,u(n-L+1)，构成当前时刻n的自适应滤波器输入向量U(n)，U(n)＝[u(n),u(n-1)...,u(n-L+1)]T，其中，L表示自适应滤波器的长度，L＝512或1024，上标T代表转置；The values u(n), u(n-1)..., u(n-L+1) of the discrete value u(n) of the input signal at the time from n to n-L+1 constitute the self-portrait of the current time n. Adaptive filter input vector U(n), U(n)=[u(n),u(n-1)...,u(n-L+1)] T , where L represents the adaptive filter The length of , L=512 or 1024, the superscript T stands for transposition; 将输入信号离散值u(n)输入自适应滤波器，得到当前时刻n的自适应滤波器的输出信号y(n)，y(n)＝WT(n)U(n)；其中，W(n)为当前时刻n自适应滤波器的权值向量，其长度等于L，初始值为零向量；Input the discrete value u(n) of the input signal into the adaptive filter to obtain the output signal y(n) of the adaptive filter at the current time n, y(n)=W T (n)U(n); where W (n) is the weight vector of the n adaptive filter at the current moment, its length is equal to L, and the initial value is a zero vector; C、回声消除C, echo cancellation 将步骤A中得到的当前时刻n的期望信号离散值d(n)，减去当前时刻n的自适应滤波器的输出信号y(n)，得到当前时刻n的误差信号e(n)，即e(n)＝d(n)-y(n)；并将当前时刻n的误差信号e(n)，作为消除回声后的纯净信号传送给远端；Subtract the output signal y(n) of the adaptive filter at the current time n from the discrete value d(n) of the desired signal at the current time n obtained in step A to obtain the error signal e(n) at the current time n, that is, e(n)=d(n)-y(n); and the error signal e(n) at the current moment n is transmitted to the far end as a pure signal after echo cancellation; D、自适应滤波器误差向量计算D. Adaptive filter error vector calculation D1、用当前时刻n到时刻n-P+1的自适应滤波器输入向量U(n),U(n-1),...,U(n-P+1)，构成自适应滤波器在当前时刻n的输入矩阵
Figure FDA0003503873340000011
Figure FDA0003503873340000012
其中，P为仿射投影阶数，其取值范围为2～6；D1. Use the adaptive filter input vectors U(n), U(n-1),...,U(n-P+1) from the current time n to the time n-P+1 to form an adaptive filter the input matrix at the current time n
Figure FDA0003503873340000011
Figure FDA0003503873340000012
Among them, P is the affine projection order, and its value range is 2～6; D2、用当前时刻n到时刻n-P+1的期望信号d(n),d(n-1),...,d(n-P+1)，构成当前时刻n的期望信号向量D(n)，D(n)＝[d(n),d(n-1),...,d(n-P+1)]T；D2. Use the expected signals d(n), d(n-1),...,d(n-P+1) from the current time n to the time n-P+1 to form the expected signal vector D of the current time n (n), D(n)=[d(n),d(n-1),...,d(n-P+1)] T ; D3、将当前时刻n的期望信号向量D(n)减去当前时刻n的输入矩阵
Figure FDA0003503873340000021
的转置与自适应滤波器的权值向量W(n)的乘积，得到当前时刻n的自适应滤波器的误差向量E(n)，
Figure FDA0003503873340000022
D3. Subtract the input matrix of the current time n from the expected signal vector D(n) at the current time n
Figure FDA0003503873340000021
The product of the transpose of and the weight vector W(n) of the adaptive filter, the error vector E(n) of the adaptive filter at the current moment n is obtained,
Figure FDA0003503873340000022
 E、自适应滤波器抽头权值更新E, adaptive filter tap weight update E1、计算当前时刻n的权值误差向量V(n)，V(n)＝Wo-W(n)，其中，Wo为自适应滤波器的权值向量的期望值，也即待估计的回声信道；再计算当前时刻n的权值误差矩阵Q(n)，Q(n)＝V(n)V(n)T；E1. Calculate the weight error vector V(n) of the current moment n, V(n)=W o -W(n), where W o is the expected value of the weight vector of the adaptive filter, that is, the value to be estimated Echo channel; recalculate the weight error matrix Q(n) of the current moment n, Q(n)=V(n)V(n) T ; E2、混合参数的计算E2. Calculation of mixing parameters 计算当前时刻n的权值误差矩阵迹J(n)，J(n)＝Tr(Q(n))，其中，Tr(·)表示求矩阵的迹；Calculate the weight error matrix trace J(n) of the current moment n, J(n)=Tr(Q(n)), where Tr( ) represents the trace of the matrix; 利用步骤B中的滤波器长度L，步骤D1中的输入矩阵
Figure FDA0003503873340000023
和仿射投影阶数P，计算当前时刻n的混合参数λ(n)：Using the filter length L in step B, the input matrix in step D1
Figure FDA0003503873340000023
and the affine projection order P, calculate the mixing parameter λ(n) of the current moment n:
Figure FDA0003503873340000024
Figure FDA0003503873340000024
 其中，μ2是小步长，μ1是大步长，两者取值范围均是0.0001～0.5，
Figure FDA0003503873340000025
是回声信号的方差，取值范围为0.001～0.1；Among them, μ 2 is a small step size, μ 1 is a large step size, and the value range of both is 0.0001～0.5,
Figure FDA0003503873340000025
is the variance of the echo signal, ranging from 0.001 to 0.1; E3、复位机制的触发E3. Trigger of reset mechanism 计算当前时刻n的信号误差的相对强度
Figure FDA0003503873340000026
同时计算前一时刻n-1的信号误差绝对强度r(n-1)＝log10(e2(n-1))；Calculate the relative strength of the signal error at the current time n
Figure FDA0003503873340000026
Simultaneously calculate the absolute intensity of the signal error at the previous moment n-1 r(n-1)=log 10 (e 2 (n-1)); 如果R(n)≤t1，或者R(n)＞t1且r(n-1)≥t2，则判定当前时刻回声信道没有发生突变，进行E4步操作；其中，t1表示信号误差的相对强度的阈值，取值范围是1～3，t2表示信号误差绝对强度的阈值，取值范围是-3～-1，q0表示复位初始值，取值范围是1～10；If R(n)≤t 1 , or R(n)>t 1 and r(n-1)≥t 2 , it is determined that the echo channel has no sudden change at the current moment, and the operation of step E4 is performed; wherein, t 1 represents the signal error The threshold value of the relative strength of , the value range is 1~3, t 2 represents the threshold value of the absolute strength of the signal error, the value range is -3~-1, q 0 represents the reset initial value, and the value range is 1~10; 如果R(n)＞t1同时r(n-1)＜t2，则判定当前时刻回声信道发生突变，对权值误差矩阵迹进行复位，即令J(n)＝q0；同时将E2步骤中混合参数λ(n)计算公式中的J(n)替换为q0，得到复位后的当前时刻n的混合参数λ(n)；If R(n)>t 1 and r(n-1)<t 2 , it is determined that the echo channel has a sudden change at the current moment, and the weight error matrix trace is reset, that is, J(n)=q 0 ; Replace J(n) in the calculation formula of the mixing parameter λ(n) with q 0 to obtain the mixing parameter λ(n) of the current moment n after reset; E4、滤波器抽头权值的更新E4. Update of filter tap weights 利用步骤D1中的输入矩阵
Figure FDA0003503873340000031
步骤D3中的误差向量E(n)以及当前时刻n的混合参数λ(n)，得到下一时刻(n+1)的滤波器抽头权系数W(n+1)：Utilize the input matrix from step D1
Figure FDA0003503873340000031
The error vector E(n) in step D3 and the mixing parameter λ(n) of the current moment n are used to obtain the filter tap weight coefficient W(n+1) of the next moment (n+1):
Figure FDA0003503873340000032
Figure FDA0003503873340000032
 其中，(·)-1表示求逆矩阵；Among them, ( ) -1 represents the inverse matrix; F、权值误差矩阵迹的更新F. Update of weight error matrix trace 利用步骤B中的滤波器长度L，步骤D1中的输入矩阵
Figure FDA0003503873340000033
和仿射投影阶数P，步骤E2中的混合参数λ(n)和权值误差矩阵Q(n)，得到下一时刻(n+1)的权值误差矩阵迹J(n+1)：Using the filter length L in step B, the input matrix in step D1
Figure FDA0003503873340000033
and the affine projection order P, the mixing parameter λ(n) and the weight error matrix Q(n) in step E2, to obtain the weight error matrix trace J(n+1) at the next moment (n+1):
Figure FDA0003503873340000034
Figure FDA0003503873340000034
 G、重复G. to repeat 令n＝n+1，重复步骤A、B、C、D、E、F，直至通话结束。Let n=n+1, repeat steps A, B, C, D, E, F until the call ends.
CN201810547634.7A 2018-05-31 2018-05-31 A Combined Step-Size Echo Cancellation Method with High Tracking Performance Active CN108877824B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810547634.7A CN108877824B (en) 2018-05-31 2018-05-31 A Combined Step-Size Echo Cancellation Method with High Tracking Performance

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810547634.7A CN108877824B (en) 2018-05-31 2018-05-31 A Combined Step-Size Echo Cancellation Method with High Tracking Performance

Publications (2)

Publication Number Publication Date
CN108877824A CN108877824A (en) 2018-11-23
CN108877824B true CN108877824B (en) 2022-03-25

Family

ID=64336595

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810547634.7A Active CN108877824B (en) 2018-05-31 2018-05-31 A Combined Step-Size Echo Cancellation Method with High Tracking Performance

Country Status (1)

Country Link
CN (1) CN108877824B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110572525B (en) * 2019-10-30 2021-05-07 西南交通大学 An adaptive communication echo cancellation method for voice communication
CN111050005B (en) * 2019-12-19 2021-02-09 四川大学 Bias compensation collective affine projection echo cancellation method
CN112397080B (en) * 2020-10-30 2023-02-28 浙江大华技术股份有限公司 Echo cancellation method and apparatus, voice device, and computer-readable storage medium
CN117896467B (en) * 2024-03-14 2024-05-31 苏州大学 Echo cancellation method and system for stereo telephone communication

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6240133B1 (en) * 1998-02-05 2001-05-29 Texas Instruments Incorporated High stability fast tracking adaptive equalizer for use with time varying communication channels
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
US9036815B2 (en) * 2012-06-02 2015-05-19 Yuan Ze University Method for acoustic echo cancellation and system thereof
CN103369162B (en) * 2013-07-09 2015-08-05 西南交通大学 A kind of listener's echo self adaptive elimination method of low complex degree
CN104410761B (en) * 2014-09-13 2016-03-02 西南交通大学 A kind of affine projection symbol subband convex combination adaptive echo cancellation method
CN107105111B (en) * 2017-03-15 2019-08-02 西南交通大学 A kind of proportional affine projection echo cancel method of combination step-length
WO2019051841A1 (en) * 2017-09-18 2019-03-21 深圳市汇顶科技股份有限公司 Method for determining filter coefficient and device therefor, and terminal

Also Published As

Publication number Publication date
CN108877824A (en) 2018-11-23

Similar Documents

Publication Publication Date Title
CN109754813B (en) Variable step size echo cancellation method based on rapid convergence characteristic
CN108877824B (en) A Combined Step-Size Echo Cancellation Method with High Tracking Performance
CN110572525B (en) An adaptive communication echo cancellation method for voice communication
CN104601837B (en) A kind of robust convex combination self adaptation listener&#39;s echo removing method
CN103369162B (en) A kind of listener&#39;s echo self adaptive elimination method of low complex degree
CN113873090B (en) Robust estimation affine projection spline self-adaptive echo cancellation method
CN111028856A (en) Echo cancellation method with variable step length
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
Huang et al. Practically efficient nonlinear acoustic echo cancellers using cascaded block RLS and FLMS adaptive filters
CN107819963B (en) A kind of equal cube absolute value echo cancel method of the minimum of convex combination
CN105721729A (en) Echo cancellation method based on block-sparse proportionate reuse weight coefficient affine projection
CN107333017B (en) An Adaptive Echo Cancellation Method Using Reuse Weight Exponential Error
CN113409806A (en) Zero attraction echo cancellation method based on arc tangent function
CN105788605B (en) A kind of affine projection symbol echo cancel method based on orthogonal transformation
CN109040497A (en) A kind of proportional class illumination-imitation projection self-adoptive echo cancel method based on M estimation
CN111739548B (en) A Minimum Error Entropy Adaptive Echo Cancellation Method Against Input Noise
CN106060295B (en) A kind of proportional affine projection echo cancel method of convex combination coefficient difference
CN107610714A (en) An Echo Cancellation Method Based on the Absolute Value of Least Mean Cubic Power of One Norm Zero Attraction
CN106161823A (en) Based on the sub band structure echo cancel method mixing P norm continuously
CN109697986A (en) A kind of adaptive deviation compensation echo cancel method based on minimum cube absolute value
CN118038844A (en) Conference scene echo cancellation method based on adaptive filtering
CN110767245B (en) Voice communication self-adaptive echo cancellation method based on S-shaped function
Imen et al. The NP-VSS NLMS algorithm with noise power estimation methods for acoustic echo cancellation
CN106161822A (en) A kind of normalization sub-band adaptive echo cancel method estimated based on M

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
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20240811

Address after: 518000 1104, Building A, Zhiyun Industrial Park, No. 13, Huaxing Road, Henglang Community, Longhua District, Shenzhen, Guangdong Province

Patentee after: Shenzhen Hongyue Information Technology Co.,Ltd.

Country or region after: China

Address before: 610031 No. two, section 111, ring road, Chengdu, Sichuan, China

Patentee before: SOUTHWEST JIAOTONG University

Country or region before: China