58results about How to "Good echo cancellation" patented technology

The invention discloses an adaptive echo cancellation method adopting memory proportionate affine projection and based on M-estimation. The adaptive echo cancellation method comprises the following steps: A, a remote signal is sampled; B, in echo estimation, a filter input vector X(n) passes through an adaptive filter and an output value y(n), namely an echo estimation value y(n), is obtained and equal to XT(n)w(n); C, in echo cancellation, a near-end signal d(n) with the echo, acquired by a near-end microphone and the output value y(n) of the adaptive filter are subjected to subtraction operation, and a return signal, namely a residual signal e(n), is returned to a far end and equal to d(n)-y(n); D, in updating of a tap weight vector of the filter, the method adopting memory proportionate affine projection and based on M-estimation is used and the tap weight vector w(n+1) of the adaptive filter at next moment n+1 is calculated and equal to w(n)+mu P(n)(UT(n)P(n)+delta IK)-1 psi [E(n)]; E, n is set to be n+1, and the steps B, C and D are repeated until a call is ended. The method has a good cancellation effect on an acoustic echo of a communication system and is high in convergence speed and small in steady-state error.

An improved convex combination decorrelation proportionate self-adaption echo cancellation method comprises the steps that first, far-end signal filtering is carried out, the input vector X(n) of a convex combination self-adaption echo cancellation filter is formed by the discrete value of a far-end signal, and after filtering is carried out on the input vector, a large-step-length filtering value y1(n) and a small-step-length filtering value y2(n) are obtained; second, decorrelation operation is carried out on the input vector X(n), the result of decorrelation operation serves as the weight coefficient updating direction vector Z(n) of the convex combination self-adaption echo cancellation filter; third, convex combination is carried out, the large-step-length filtering value y1(n) and the small-step-length filtering value y2(n) are subjected to convex combination through weight lambda (n), and a combination filter value y(n) is obtained; fourth, echo cancellation is carried out, the combination filter value y(n) is subtracted from a near-end signal d(n) with an echo, and the subtracted near-end signal d(n) is fed back to the far end; fifth, a filter tap weight coefficient is updated; sixth, the weight of the filter is updated; seventh, the weight of the filter is limited; eighth, n is made to be equal to n+1, the first step to the seventh step are repeated till a conversation is over. According to the method, a high rate of convergence can be obtained, the steady state error can be small, and a good anti-jamming capability is achieved.

The invention discloses a convex combination adaptive echo cancellation method for an affine projection sign subband adaptive filter. The method mainly comprises the following steps: A, far-end signal filtering: performing far-end signal filtering to obtain the output vector Y1(n) of a large-step-length filter and the output vector Y2(n) of a small-step-length filter; B, convex combination: performing convex combination on the output vector Y1(n) and the output vector Y2(n) of the large-step-length filter and the small-step-length filter to obtain a combination filtering value Y(n) (Y(n)= lambda(n)Y1(n)+(1-lambda(n))Y2(n)), and performing convex combination on the tap weight vector W1(n) of the large-step-length filter and the tap weight vector W2(n) of the small-step-length filter to obtain a total filter group tap weight vector W(n) (W(n)=lambda(n)W1(n)+(1-lambda(n))W2(n); C, echo cancellation: subtracting the combination filter value Y(n) from an echo-carrying affine projection near-end signal D(n) picked up by a near-end microphone to obtain a net signal E(n) (E(n)=D(n)-Y(n)), and transmitting the net signal back to a far end; D, updating the tap weight coefficients of the filters; E, weight update of the filters: updating hybrid parameters a(n); and F, iteration: repeating the steps above under the condition that n=n+1 till the end of a conversation. The method has the advantages of high cancellation capability on the acoustic echo of a sparse communication system, high convergence speed, small steady-state error and good echo cancellation effect.

The invention discloses an echo cancellation method based on block-sparse proportionate reuse weight coefficient affine projection. The method mainly comprises following steps of A, filtering a far-end signal; obtaining an output value y (n) from a filter input vector X (n) through a self-adaptive filter; B, cancelling an echo; subtracting the output value y (n) by near-end signal d (n) which is picked up by a near-end microphone at a current moment and is equipped with the echo, thus obtaining an error as a residual signal e (n), wherein e (n) is equal to d (n)-y(n); sending back to a far-end; C, updating a filter tapping weight vector; D, assuming that n is equal to n+1, repeating steps A, B and C until finishing a call. According to the method, the echo elimination capability to the voice communication of a sparse system is high; the rate of convergence is high; the steady state error is low; and the echo elimination effect is good.

The invention provides a robust convex combination type adaptive phone echo canceling method. The method comprises the main steps of A, filtering a far-end signal to obtain quick adaptive filter output y1(n) and low-error filter output y2(n); B, convex combination, namely, performing convex combination for the outputs of the two filters to obtain output y(n) of convex combination filters; c, canceling echo, namely, receiving a near-end signal d(n) with echo by a near-end microphone, subtracting y(n) and then transmitting to a far end as an echo-canceled near-end signal e(n); D, updating tap weight coefficients of the filters; E, updating the weights of the filter, and updating the hybrid parameter a (n) through a formula with sign functions; F, limiting the weight of the filters; G, enabling n to be n+1, and repeating A, B, C, D, E and F until the call is ended. The method is high in convergence rate, low in steady-state error, high in interference resistance, good in echo canceling effect, and maintains good working state under the condition of double-talk of a phone.

The present application discloses an echo cancellation method and device. The echo cancellation method includes: updating a parameter of a cloud echo cancellation system by using a first reference signal at a first moment, wherein the first reference signal is a voice synthesizing result after the cloud server responds to a first user instruction; using a target reference signal to perform echo cancellation on a microphone signal through the cloud echo cancellation system at a second moment after the first moment if the microphone signal and the target reference signal uploaded by the client are received. Since cloud echo cancellation can use a more complex echo cancellation algorithm than local and pre-estimate echo cancellation filter parameters, a better echo cancellation effect can beachieved when echo cancellation is performed on the microphone signal.

A zero-norm set membership affine projection adaptive echo cancellation method based on weight vector reuse comprises the following steps: A, remote signal sampling; B echo estimation: obtaining an output value y(n) at a current time n through an adaptive filter based on an input vector X(n) of the adaptive filter at the current time n, wherein y(n) is the estimated value of echo, and y(n)=X<T>(n)*w(n); C, echo cancellation: subtracting the output value y(n) of the adaptive filter at the current time n from a near-end signal d(n) with echo at the current time n picked up by a near-end microphone to get a residual signal e(n), wherein e(n)=d(n)-y(n), and transmitting the residual signal back to a remote end; D, filter tap weight vector updating: getting a tap weight vector w(n+1) (Please see the description for the expression) of the filter at next time n+1 using a zero-norm set membership affine projection method based on weight vector reuse; and E, letting n=n+1, and repeating the steps B, C and D until the end of the call. The method has the advantages of good cancellation effect for acoustic echo of a communication system, high convergence rate, and small steady-state error.

A method for eliminating combined step length proportional affine projection echo comprises the following steps of A, acquiring a sound signal x(n) transmitted from a far end at a telephone loudspeaker and acquiring an expected signal d(n) at a telephone microphone; B, using the sound signal x(n) as an input of an adaptive filter, and using the output of the adaptive filter as an estimated value d<^>(n) of an echo signal; C, performing echo offset, namely using an error signal e(n) obtained through subtracting d<^>(n) from d(n) as a clean signal after echo offset and transmitting the error signal to the remote end; D, calculating an error vector of the adaptive filter; E, calculating a weight updating increment deltaw(n) of the adaptive filter; F, calculating a combined step length mu(n); G, updating the weight vector w(n+1)=w(n)+mu(n)deltaw(n) of the adaptive filter; and H, setting n=n+1, repeating the steps A, B, C, D, E, F, G until the calling is finished. The method has advantages of high convergence speed and low steady-state error.

An adaptive echo cancellation method based on correlation entropy is disclosed. The method comprises the following steps: A, collecting far-end signals, sampling far-end signals transmitted from far-end to obtain the discrete value x (n) of far-end input signals at the current time n, and the filter input signal vector is x (n) = [x (n), x (n-1),..., x (n-L+1)] T; B, sampling far-end signals transmitted from far-end to obtain the discrete value x (n) of far-end input signals at the current time n. B, estimate an echo signal, passing an input signal vector x (n) of that current time n through an adaptive filt, and outputting a value y (n) thereof, that is, an estimated value of the echo signal; C, canceling the echo, sampling with a near-end microphone to obtain the near-end signal d (n) ofthe current time n of the echo return, subtracting the estimated value y (n) of the echo signal; D, updating the tap weight coefficients of the filter, and calculating a tap weight vector w (n+1), w(n+1) = w (n) + Mu(n) U (n) (UT (n) U (n)) of n+1 at the next time of the filter; 1E(n)-C (n); E, let n=n+1, repeating the process of A, B, C, D, E until the end of the call. This method can obtain faster convergence rate and lower steady-state error, and the echo cancellation effect is good.

The invention discloses a self-adaptive echo cancellation method based on time-varying parameters. The adaptive echo cancellation method comprises the following steps: A, far-end signal sampling; B, echo signal estimation: obtaining an output value y (n) from the filter input vector X (n) through an seld-adaptive filter, namely, an echo estimation value y (n), y (n) = WT (n) X (n); C, echo signalelimination: subtracting the near-end signal d (n) with echo picked up by the near-end microphone from the output value y (n) of the self-adaptive filter, and sending the subtracted near-end signal d(n) and the subtracted output value y (n) back to the far end, the returned signal being a residual signal e (n), and e (n) = d (n)-; y (n); D, updating a tap weight vector of the filter; estimating aresidual signal square sequence Ae (n) in the window according to the current moment n, obtaining a smooth square value of the residual signal, further obtaining an adjusting parameter (n) of a time-varying proportional control factor and a time-varying proportional control factor gk (n), and finally obtaining an adaptive filter tap weight W (n + 1) at the next moment; and E, enabling n to be equal to n + 1, and repeating the steps A, B, C and D until the conversation is finished. The method has the advantages of good elimination effect on acoustic echoes of a communication system, high convergence speed and small steady-state error.

The present invention discloses an affine projection symbol echo cancellation method based on orthogonal transformation. The method mainly comprises the steps of A acquiring a voice signal x(n) transmitted from a far-end and a desired signal d(n) collected by a near-end microphone; B calculating the output y(n) of an adaptive filter; C carrying out the echo cancellation, namely, taking an error signal e(n) obtained by subtracting y(n) from d(n) as a clean signal after the echo cancellation to transmit to the far-end; D orthogonally transforming an input matrix X(n) of the adaptive filter into a decorrelation matrix U(n); E updating a tap weight vector of the adaptive filter; F defining n=n+1, and repeating the steps of A, B, C, D and E until the conversation is ended. The method has the advantages of fast convergence speed and low steady-state error except having the good robustness for the doubletalk.

The invention discloses a proportional affine projection echocancellation method of convex combination coefficient differences. The method includes the following steps: A. far-end signals sampling and filtering, A1. a far-end dispersion input signal constitutes an input vector X (n) of a convex combination adaptive filtering echocancellation filter; A2. inputting a filter to the vector X (n) to obtain filtering values y1 (n) and y2 (n) of a large step length and a small step length; B. echo offsetting, subtracting a near-end signal d (n) and a combination filter output value y (n) to obtain a total residual signal e (n) after echocancellation, and transmitting the total residual signal e (n) to a far end; C. conducting convex combination, conducting convex combination on the output value y1 (n) and y2 (n) of the large step length and the small step length through a weight [lambda] to obtain a combination output value y (n), and conducting convex combination on tap weight coefficients W1 (n) and W2 (n) through the weight [lambda] to obtain a tap weight coefficient W (n) of a tap filter; D. updating a tap weight vector of a filter; E. updating the weight of the filter; F. limiting the weight of the filter; G. repeating the steps of A, B, C, D, E, F, thus realizing echocancellation. The method has strong identification capability, rapid convergence, strong tracking capability, low stability error rate, and excellent effects of echocancellation.

A microphone array device including microphone capsules and at least one processing unit configured to receive output signals of the microphone capsules, dynamically steer an audio beam based on the received output signal of the microphone capsules, and generate and provide an audio output signal based on the received output signal of the microphone capsules. The processing unit is configured to operate in a dynamic beam mode where at least one focused audio beam is formed that points towards a detected audio source and in a default beam mode where a broader audio beam is formed that covers substantially a default detection area. The microphone array may be incorporated into a conference system.

The invention provides an echo cancellation method and device, electronic equipment and a readable storage medium, and aims to improve the echo cancellation effect. The method comprises the following steps: acquiring a to-be-processed mixed voice signal and a to-be-processed far-end voice signal; inputting the to-be-processed mixed voice signal into a pre-trained network model to obtain a first feature vector; inputting a far-end voice signal to be processed into the network model to obtain a second feature vector; inputting the first feature vector and the second feature vector into a pre-trained attention model to obtain a third feature vector; and obtaining a near-end voice signal after echo cancellation according to the first feature vector and the third feature vector. The echo cancellation method and device, the electronic equipment and the readable storage medium provided by the invention have a better echo cancellation effect.