Echo Cancellation Method Based on Adaptive Decorrelation and Variable Step-Size Proportional M Estimation

Abstract

The present invention discloses an echo cancellation method based on adaptive decorrelation and variable step-size proportional M estimation, comprising the following steps: A. signal acquisition; B. calculating the decorrelation coefficient vector #imgabs0# according to the speech signal sampling and the expected signal sampling results; C. calculating the decorrelation input vector x D (n) and the expected signal d after decorrelation D (n); D. Calculate the output y(n) of the adaptive filter and the output y of the adaptive filter after decorrelation D (n); E, calculate the error signal e(n) and the error signal e after decorrelation D (n); F, calculate the proportional matrix G(n); G, calculate the error signal #imgabs1# after robust decorrelation for iteration in M ​​estimation; H, calculate the step size parameter μ(n); I, update the tap weight vector of the adaptive filter and enter the processing at the next moment. In addition to having good robustness to impulse noise, the present invention also has the advantages of fast convergence speed and good steady-state error.

Description

Technical Field

[0001] The invention belongs to the technical field of adaptive echo cancellation of voice communication, and in particular relates to an echo cancellation method based on adaptive decorrelation and variable step-size proportional M estimation. Background Art

[0002] In the process of voice communication, acoustic echo is inevitable wherever a microphone and a speaker are used at the same time. The voice of the far-end speaker is played through the near-end speaker, directly or indirectly received by the near-end microphone, and transmitted back to the far-end, so that the far-end speaker can hear his own delayed voice. This is acoustic echo. The propagation path of sound from the speaker to the microphone is called the echo path, and its impulse response vector is recorded as w o The impulse response of the acoustic echo path is often sparse, i.e., w o Most of the elements are zero or close to zero, and a small number of elements have large amplitudes. Acoustic echo is the most important factor affecting the quality of voice calls.

[0003] At present, in order to eliminate echo, the most effective technology recognized internationally is adaptive echo cancellation technology. In essence, adaptive speech echo cancellation is also a problem of identifying the impulse response of the echo path, that is, the adaptive filter can adjust the adaptive filter weights (also the estimated value of the echo path impulse response) according to the changes in the environment, obtain the estimated value of the speech echo (the output signal of the adaptive filter), and then subtract the estimated value from the signal received by the near-end microphone to obtain a clean signal and transmit it to the far end to achieve the purpose of eliminating the echo. Therefore, the key issue is to design an adaptive filter algorithm with excellent performance. Due to the sparseness of the echo path impulse response, the proportional normalized least mean square (PNLMS) algorithm has a faster convergence speed than the normalized least mean square (NLMS) algorithm. This is because the PNLMS algorithm can use the prior condition of channel sparseness to allocate more gain to the filter coefficients with large amplitude to achieve rapid convergence, thereby improving the overall convergence speed of the algorithm.

[0004] However, in actual calls, speech signals are often affected by impact noise. In this case, the performance of the PNLMS algorithm will decline or even diverge. Based on the Normalized least mean M-estimate (NLMM) algorithm, Huang Zhangliang combined the PNLMS algorithm and used the improved Huber norm to obtain a type of proportional normalized least mean M-estimate (PNLMM) algorithm, which has better resistance to impact noise. "Huang Zhangliang. Proportional Adaptive Robust Echo Cancellation Algorithm [D]. Chengdu: Southwest Jiaotong University, 2012." However, since the speech signal is a highly correlated and non-stationary signal, and the PNLMM algorithm has no decorrelation capability. Therefore, in speech echo cancellation, the convergence speed of the PNLMM algorithm is not ideal. On the other hand, since the PNLMM algorithm uses a fixed step size, there is a contradiction in the trade-off between convergence speed and convergence level. That is, when the step size is large, the steady state of the algorithm is poor and the convergence speed is fast; conversely, when the step size is small, the convergence speed of the algorithm is slow and the convergence level is good. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide an echo cancellation method based on adaptive decorrelation and variable step-size proportional M estimation, which has strong decorrelation ability for speech signals, fast convergence speed, good robustness and good echo cancellation effect.

[0006] The object of the present invention is achieved through the following technical scheme: an echo cancellation method based on adaptive decorrelation and variable step-size proportional M estimation comprises the following steps:

[0007] A. Signal Acquisition

[0008] The speech signal transmitted from the far end and played through the near-end speaker is sampled to obtain the discrete value x(n) of the far-end sound signal at the current time n. At the same time, the expected signal collected by the near-end microphone is sampled to obtain the discrete value d(n) of the expected signal at the current time n (where d(n) includes the echo signal and interference noise);

[0009] B. Calculate the decorrelation coefficient vector

[0010] B1. The values ​​of the far-end audio signal discrete values ​​x(n) obtained in step A at the time n-1 to nK constitute the adaptive decorrelation input vector u(n) at the current time n, u(n) = [x(n-1) x(n-2) … x(nK)] T, where the superscript T represents the transposition operation of the vector, K represents the decorrelation order, K ≥ 1. As the value of K increases, the convergence speed of the algorithm will increase, but the steady state will deteriorate;

[0011] B2. Calculate the adaptive decorrelation error signal e at the current time n xd (n), in is the tap weight vector of the adaptive decorrelator at time n-1, its length is equal to K, and its initial value is a zero vector;

[0012] B3. Update the adaptive decorrelation coefficient vector at time n The calculation formula is:

[0013]

[0014] where μ a is the decorrelation step length, ranging from 0.001 ≤ μ a ≤0.05;

[0015] C. Calculate the decorrelated input vector x D (n) and the expected signal d after decorrelation D (n)

[0016] C1. Calculate the decorrelated input signal x at the current time n D (n), Use the decorrelated input signal x D (n) The values ​​from time n to time n-L+1 constitute the decorrelated input vector x at the current time n. D (n), x D (n)=[x D (n) x D (n-1) … x D (n-L+1)] T , where L is the number of taps of the adaptive filter, which is usually L=512 or 1024 in acoustic echo cancellation;

[0017] C2. Use the values ​​of the expected signal discrete values ​​d(n) obtained in step A from time n-1 to time nK to form the decorrelation expected vector d(n) at the current time n, d(n)=[d(n-1) d(n-2) … d(nK)] T ; Calculate the expected signal d after decorrelation at the current time n D (n),

[0018] D. Calculate the output y(n) of the adaptive filter and the output y after decorrelation D (n)

[0019] D1. Use the values ​​of the remote audio signal discrete values ​​x(n) obtained in step A from time n to time n-L+1 to form the input vector x(n) at the current time n, x(n)=[x(n) x(n-1) … x(n-L+1)] T , calculate the output signal y(n) of the adaptive filter at the current time n, y(n)=w T (n)x(n), where w(n)=[w1(n) w2(n) … w L (n)] T is the tap weight vector of the adaptive filter at time n, its length is equal to L, and its initial value is a zero vector;

[0020] D2, using the decorrelated input vector x at the current time n in step C1 D (n), calculate the decorrelated adaptive filter output y at the current time n D (n), y D (n)=w T (n)x D (n);

[0021] E. Calculate the error signal e(n) and the error signal e after decorrelation D (n)

[0022] E1, subtract the adaptive filter output signal y(n) obtained in step D1 from the near-end expected signal d(n) at the current time n obtained in step A, and obtain the error signal e(n) at time n, that is, e(n)=d(n)-y(n), and transmit e(n) to the far-end as the clean signal after echo elimination, so that the far-end speaker cannot hear his previous voice, thereby achieving the purpose of echo elimination;

[0023] E2, the decorrelated expected signal d at the current time n obtained in step C2 D (n) Subtract the output y of the decorrelated adaptive filter obtained in step D2 D (n), and obtain the decorrelated error signal e at time n D (n), that is, e D (n) = d D (n)-y D (n);

[0024] F. Calculate the proportional matrix G(n)

[0025] The proportional matrix G(n) is a diagonal matrix G(n) = diag[g1(n) g2(n) … g L (n)], where the pth diagonal element g at the current moment n p (n) by Calculate, 1≤p≤L, where κ is usually 0 or -0.5 or -0.75, and ||·||1 represents the 1-norm of the vector;

[0026] G. Calculate the error signal after robust decorrelation for iteration in M ​​estimation

[0027] G1, using the decorrelated error signal e at time n obtained in step E2 D (n) squared From time n to time nN w The value of +1 constitutes the error estimate sample A after decorrelation at the current moment e (n), Where N w is the sample length of the error estimate after decorrelation selected;

[0028] Calculate the error variable σ after decorrelation without shock interference 2 (n), σ 2 (n) = λσ 2 (n-1)+C(1-λ)med(A e (n)), where σ 2 The initial value of (n) is zero, the λ forgetting factor, and 0<<λ<1, C=2.2[1+5 / (N w +1)] 2 ,med(·) is the median operator;

[0029] The calculation formula of the error signal threshold parameter ξ after decorrelation is, ξ=2.576σ 2 (n);

[0030] G2, error signal after robust decorrelation The calculation formula is:

[0031] H. Calculate the step size parameter μ(n)

[0032] H1, the error signal after robust decorrelation obtained in step G2 Calculate the mean square robust error at time n Where χ is the forgetting factor, and 0<<χ<1;

[0033] H2, according to the decorrelated input signal x obtained in step C1 D (n) Calculate the mean square error of the decorrelated input signal at time n

[0034] Calculate the decorrelated input signal vector x at time n D (n) and the robust error signal The relevant parameters r(n),

[0035] Calculate the excess mean square error at time n

[0036] H3. Calculate the value of the step size parameter μ(n) at time n,

[0037] I. Update the tap weight vector of the adaptive filter and enter the processing of the next moment:

[0038] Calculate the tap weight vector w(n+1) of the adaptive filter at the next moment n+1,

[0039]

[0040] Let n=n+1, and repeat steps A, B, C, D, E, F, G, H, and I until the call is ended.

[0041] The beneficial effects of the present invention are as follows: the present invention utilizes an adaptive decorrelation method to perform decorrelation processing on a remote sound and an expected signal, obtains the decorrelated sound and the expected signal, and then sends them to an adaptive filter for adaptive processing; since direct decorrelation requires a large amount of calculation, the present invention adjusts the direct decorrelation to adaptive decorrelation to greatly reduce the computational complexity of the algorithm for high-order decorrelation, shortens the time for decorrelation processing on the collected signal, and thus shortens the overall running time of the algorithm; on the other hand, the present invention minimizes the a posteriori error and utilizes the fastest gradient descent method to derive the optimal formula for the step size, thereby achieving a large step size during the initial iteration, and the algorithm converges quickly, and then the step size gradually decreases with the iteration, thereby obtaining a better steady state. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of the method of the present invention;

[0043] Figure 2 is the impulse response of the echo path in the simulation experiment.

[0044] Figure 3 are the far-end speech signal and the near-end expected signal in the simulation experiment.

[0045] Figure 4 are the misalignment curves of algorithms PNLMS, PNLMM and 1 implementation of the present invention.

[0046] Figure 5 It is the clean signal obtained by filtering the far-end speech signal using the estimated echo path impulse response and subtracting the estimated expectation from the near-end expectation signal. DETAILED DESCRIPTION

[0047] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the following.

[0048] like Figure 1 As shown, the echo cancellation method based on adaptive decorrelation and variable step-size proportional M estimation includes the following steps:

[0049] A. Signal Acquisition

[0050] The speech signal transmitted from the far end and played through the near-end speaker is sampled to obtain the discrete value x(n) of the far-end sound signal at the current time n. At the same time, the expected signal collected by the near-end microphone is sampled to obtain the discrete value d(n) of the expected signal at the current time n (where d(n) includes the echo signal and interference noise);

[0051] B. Calculate the decorrelation coefficient vector

[0052] B1. The values ​​of the far-end audio signal discrete values ​​x(n) obtained in step A at the time n-1 to nK constitute the adaptive decorrelation input vector u(n) at the current time n, u(n) = [x(n-1) x(n-2) … x(nK)] T , where the superscript T represents the transposition operation of the vector, K represents the decorrelation order, K ≥ 1. As the value of K increases, the convergence speed of the algorithm will increase, but the steady state will deteriorate;

[0053] B2. Calculate the adaptive decorrelation error signal e at the current time n xd (n), in is the tap weight vector of the adaptive decorrelator at time n-1, its length is equal to K, and its initial value is a zero vector;

[0054] B3. Update the adaptive decorrelation coefficient vector at time n The calculation formula is:

[0055]

[0056] where μ a is the decorrelation step length, ranging from 0.001 ≤ μ a ≤0.05;

[0057] C. Calculate the decorrelated input vector x D (n) and the expected signal d after decorrelation D (n)

[0058] C1. Calculate the decorrelated input signal x at the current time n D (n), Use the decorrelated input signal x D (n) The values ​​from time n to time n-L+1 constitute the decorrelated input vector x at the current time n. D (n), x D (n)=[x D (n) x D (n-1) … x D (n-L+1)] T , where L is the number of taps of the adaptive filter, which is usually L=512 or 1024 in acoustic echo cancellation;

[0059] C2. Use the values ​​of the expected signal discrete values ​​d(n) obtained in step A from time n-1 to time nK to form the decorrelation expected vector d(n) at the current time n, d(n)=[d(n-1) d(n-2) … d(nK)] T ; Calculate the expected signal d after decorrelation at the current time n D (n),

[0060] D. Calculate the output y(n) of the adaptive filter and the output y after decorrelation D (n)

[0061] D1. Use the values ​​of the remote audio signal discrete values ​​x(n) obtained in step A from time n to time n-L+1 to form the input vector x(n) at the current time n, x(n)=[x(n) x(n-1) … x(n-L+1)] T , calculate the output signal y(n) of the adaptive filter at the current time n, y(n)=w T (n)x(n), where w(n)=[w1(n) w2(n) … w L (n)] T is the tap weight vector of the adaptive filter at time n, its length is equal to L, and its initial value is a zero vector;

[0062] D2, using the decorrelated input vector x at the current time n in step C1 D (n), calculate the decorrelated adaptive filter output y at the current time n D (n), y D (n)=w T (n)x D (n);

[0063] E. Calculate the error signal e(n) and the error signal e after decorrelation D (n)

[0064] E1, subtract the adaptive filter output signal y(n) obtained in step D1 from the near-end expected signal d(n) at the current time n obtained in step A, and obtain the error signal e(n) at time n, that is, e(n)=d(n)-y(n), and transmit e(n) to the far-end as the clean signal after echo elimination, so that the far-end speaker cannot hear his previous voice, thereby achieving the purpose of echo elimination;

[0065] E2, the decorrelated expected signal d at the current time n obtained in step C2 D (n) Subtract the output y of the decorrelated adaptive filter obtained in step D2 D (n), and obtain the decorrelated error signal e at time n D (n), that is, e D (n) = d D (n)-y D (n);

[0066] F. Calculate the proportional matrix G(n)

[0067] The proportional matrix G(n) is a diagonal matrix G(n) = diag[g1(n) g2(n) … g L (n)], where the pth diagonal element g at the current moment n p (n) by Calculate, 1≤p≤L, where κ is usually 0 or -0.5 or -0.75, and ||·||1 represents the 1-norm of the vector;

[0068] G. Calculate the error signal after robust decorrelation for iteration in M ​​estimation

[0069] G1, using the decorrelated error signal e at time n obtained in step E2 D (n) squared From time n to time nN w The value of +1 constitutes the error estimate sample A after decorrelation at the current moment e (n), Where N w is the sample length of the error estimate after decorrelation selected;

[0070] Calculate the error variable σ after decorrelation without shock interference 2 (n), σ 2 (n) = λσ 2 (n-1)+C(1-λ)med(A e (n)), where σ 2 The initial value of (n) is zero, the λ forgetting factor, and 0<<λ<1, C=2.2[1+5 / (N w+1)] 2 ,med(·) is the median operator;

[0071] The calculation formula of the error signal threshold parameter ξ after decorrelation is, ξ=2.576σ 2 (n);

[0072] G2, error signal after robust decorrelation The calculation formula is:

[0073] H. Calculate the step size parameter μ(n)

[0074] H1, the error signal after robust decorrelation obtained in step G2 Calculate the mean square robust error at time n Where χ is the forgetting factor, and 0<<χ<1;

[0075] H2, according to the decorrelated input signal x obtained in step C1 D (n) Calculate the mean square error of the decorrelated input signal at time n

[0076] Calculate the decorrelated input signal vector x at time n D (n) and the robust error signal The relevant parameters r(n),

[0077] Calculate the excess mean square error at time n

[0078] H3. Calculate the value of the step size parameter μ(n) at time n,

[0079] I. Update the tap weight vector of the adaptive filter and enter the processing at the next moment;

[0080] Calculate the tap weight vector w(n+1) of the adaptive filter at the next moment n+1,

[0081]

[0082] Let n=n+1, and repeat steps A, B, C, D, E, F, G, H, and I until the call is ended.

[0083] In the embodiments of the present application, in order to verify the effectiveness of the present invention, simulation experiments were conducted and compared with algorithms PNLMS and PNLMM in voice calls under impulse noise conditions.

[0084] 1. Simulation conditions

[0085] The impulse response of the echo path wo It was collected in a quiet and closed room with a height of 2.5m, a width of 3.75m, a length of 6.25m, a temperature of 20℃, and a humidity of 50%. Its length L is 512, such as Figure 1 As shown in the figure, the far-end speech signal x(n) and the near-end desired signal are as follows: Figure 2 As shown. The near-end microphone receives the desired signal d(n) which can be expressed by the formula d(n)=x T (n)w o +v(n) is calculated, where the observation noise v(n) is alpha stationary noise, and its characteristic function is φ(t)=exp(-γ|t| α ), where α∈(0,2] controls the characteristic index of the noise pulse characteristics, and γ>0 indicates the degree of noise dispersion. The values ​​selected in this simulation experiment are α=1.6, γ=1 / 30. The unit is decibel) and is used to evaluate the performance of each method. For fair comparison, the parameter values ​​of these algorithms are shown in Table 1.

[0086] Table 1 Parameter values ​​of each algorithm

[0087] algorithm Parameter Value PNLMS μ=0.1,κ=-0.75 PNLMM <![CDATA[μ=0.1&μ=0.5,N w =16,λ=0.98,κ=-0.75]]> The present invention <![CDATA[N w =20,λ=0.98,x=0.999,μ a =0.01,κ=-0.75,K=1]]>

[0088] Figure 2 is the impulse response of the echo path, Figure 3 The signals are collected, (a) is the far-end speech signal, and (b) is the echo signal in the near-end desired signal. Figure 4 It is the imbalance curve of the algorithm PNLMS, PNLMM and the first implementation of the present invention. Figure 3 It can be seen that under the action of impact noise, the curve of the algorithm PNLMS has more peaks, especially a maximum peak appears around the time n is 80,000, while the curves of the PNLMM algorithm and the present invention have no peaks. This shows that the present invention and the PNLMM algorithm have good robustness to impact noise. In addition, it can be seen from the figure that the present invention has a faster convergence speed and steady-state level than the PNLMS algorithm and the PNLMM algorithm. Figure 5 (a) is the estimated expected signal obtained by filtering the far-end speech signal using the estimated unknown echo path impulse response, and (b) is the clean signal obtained by subtracting the estimated expected signal from the echo signal in the near-end expected signal.

[0089] The above description shows and describes a preferred embodiment of the present invention, but as mentioned above, it should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the invention concept described herein through the above teachings or the technology or knowledge of the relevant field. And the changes and modifications made by those skilled in the art do not depart from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. An echo cancellation method based on adaptive decorrelation and variable step-size proportional M estimation, characterized in that: The following steps are involved: A. Signal acquisition: Sample the voice signal from the far end and played through the near-end speaker to obtain the discrete value x(n) of the far-end sound signal at the current time n. At the same time, sample the expected signal collected by the near-end microphone to obtain the discrete value d(n) of the expected signal at the current time n, where d(n) includes the echo signal and interference noise; B. Calculate the decorrelation coefficient vector based on the speech signal sampling and the expected signal sampling results The step B comprises the following sub-steps: B1. The values ​​of the far-end audio signal discrete values ​​x(n) obtained in step A at the time n-1 to nK constitute the adaptive decorrelation input vector u(n) at the current time n, u(n) = [x(n-1) x(n-2) … x(nK)] T , where the superscript T represents the transpose operation of the vector, K represents the decorrelation order, K ≥ 1, and as the value of K increases, the convergence speed of the algorithm will increase, but the steady state will deteriorate; B2. Calculate the adaptive decorrelation error signal e at the current time n xd (n), in is the tap weight vector of the adaptive decorrelator at time n-1, its length is equal to K, and its initial value is a zero vector; B3. Update the adaptive decorrelation coefficient vector at time n The calculation formula is: where μ a is the decorrelation step length, ranging from 0.001 ≤ μ a ≤0.05; C. Calculate the decorrelated input vector x D (n) and the expected signal d after decorrelation D (n); The step C comprises the following sub-steps: C1. Calculate the decorrelated input signal x at the current time n D (n), Use the decorrelated input signal x D (n) The values ​​from time n to time nL+1 constitute the decorrelated input vector x at the current time n D (n), x D (n) = [x D (n) x D (n-1) … x D (n-L+1)] T , where L is the number of taps of the adaptive filter; C2. Use the values ​​of the expected signal discrete values ​​d(n) obtained in step A from time n-1 to time nK to form the decorrelation expected vector d(n) at the current time n, d(n) = [d(n-1) d(n-2) … d(nK)] T ; Calculate the expected signal d after decorrelation at the current time n D (n), D. Calculate the output y(n) of the adaptive filter and the output y after decorrelation D (n); Described step D comprises the following sub-steps: D1. Use the values ​​of the remote audio signal discrete values ​​x(n) obtained in step A from time n to time n-L+1 to form the input vector x(n) at the current time n, x(n) = [x(n) x(n1) … x(nL+1)] T , calculate the output signal y(n) of the adaptive filter at the current time n, y(n) = w T (n)x(n), where w(n)=[w1(n) w2(n) … w L (n)] T is the tap weight vector of the adaptive filter at time n, its length is equal to L, and its initial value is a zero vector; D2, using the decorrelated input vector x at the current time n in step C1 D (n), calculate the decorrelated adaptive filter output y at the current time n D (n), y D (n) = w T (n)x D (n); E. Calculate the error signal e(n) and the error signal e after decorrelation D (n); e(n)=d(n)-y(n); e D (n)=d D (n)-y D (n); F. Calculate the proportional matrix G(n); The ratio matrix G(n) is a diagonal matrix G(n) = diag[g1(n) g2(n) … g L (n)], where the pth diagonal element g at the current moment n p (n) by Calculate, 1≤p≤L, where κ is usually 0 or -0.5 or -0.75, and ||·||1 represents the 1-norm of the vector; G. Calculate the error signal after robust decorrelation for iteration in M ​​estimation G1. Step G includes the following sub-steps: The decorrelated error signal e at time n D (n) squared From time n to time nN w The value of +1 constitutes the error estimate sample A after decorrelation at the current moment e (n), Where N w is the sample length of the error estimate after decorrelation selected; Calculate the error variable σ after decorrelation without shock interference 2 (n), σ 2 (n) = λσ 2 (n-1)+C(1-λ)med(A e (n)), where σ 2 The initial value of (n) is zero, the λ forgetting factor, and 0<<λ<1, C=2.2[1+5 / (N w +1)] 2 ,med(·) is the median operator; G2, error signal after robust decorrelation The calculation formula is: H. Calculate the step size parameter μ(n); The step H comprises the following sub-steps: H1, according to the error signal Calculate the mean square robust error at time n Where χ is the forgetting factor, and 0<<χ<1; H2, according to the decorrelated input signal x obtained in step C1 D (n) Calculate the mean square error of the decorrelated input signal at time n Calculate the decorrelated input signal vector x at time n D (n) and the robust error signal The relevant parameters r(n), Calculate the excess mean square error at time n H3. Calculate the value of the step size parameter μ(n) at time n, I. Update the tap weight vector of the adaptive filter and enter the processing at the next moment; Calculate the tap weight vector w(n+1) of the adaptive filter at the next moment n+1 based on the current moment n, then set n=n+1 and repeat steps A, B, C, D, E, F, G, H, and I until the call ends.

2. The echo cancellation method based on adaptive decorrelation and variable step-size proportional M estimation according to claim 1, characterized in that: Described step E comprises the following sub-steps: E1, subtract the adaptive filter output signal y(n) obtained in step D1 from the near-end expected signal d(n) at the current time n obtained in step A, and obtain the error signal e(n) at time n, that is, e(n)=d(n)-y(n), and transmit e(n) to the far-end as the clean signal after echo elimination, so that the far-end speaker cannot hear his previous voice, thereby achieving the purpose of echo elimination; E2, the decorrelated expected signal d at the current time n obtained in step C2 D (n) Subtract the output y of the decorrelated adaptive filter obtained in step D2 D (n), and obtain the decorrelated error signal e at time n D (n), namely: e D (n)=d D (n)-y D (n)。 3. The echo cancellation method based on adaptive decorrelation and variable step-size proportional M estimation according to claim 1, characterized in that: In step I, the tap weight vector w(n+1) of the adaptive filter at the next time n+1 is calculated according to the current time n, thereby updating the tap weight vector of the adaptive filter: Then set n=n+1 and repeat steps A, B, C, D, E, F, G, H, and I until the call is ended.

Images