Echo cancelling method and echo canceller using the same

Claim 2

An echo canceller which repeats operations of: providing a received signal onto an echo path to obtain therefrom an echo; generating an echo replica by applying said received signal to estimated echo path means formed by a digital adaptive filter having an estimated impulse response as a filter coefficient; subtracting said echo replica by subtracting means from said echo and outputting the resulting error signal as an echo-cancelled or echo-free signal; and estimating said filter coefficient of said estimated echo path means, as its impulse response, by impulse response estimating means on the basis of said error signal and said received signal and iteratively adjusting said filter coefficient by a recursive least-squares (RLS) algorithm; wherein said impulse response estimating means includes: adjustment matrix storage means for storing an adjustment matrix representing an expectation of an impulse response variation of said echo path; and adding means for adding said adjustment matrix to a coefficient error covariance matrix in said RLS logarithm to generate an adjusted coefficient error covariance matrix.

Claim 3

The echo canceller of claim 1 wherein said impulse response estimating means includes: gain vector generating means for generating a gain vector on the basis of said adjusted coefficient error covariance matrix and the vector of said received signal; covariance matrix generating means for generating an updated version of said coefficient error covariance matrix on the basis of said adjusted coefficient error covariance matrix and said gain vector; multiplying means for multiplying said gain vector and said error signal to generate an adjustment coefficient; coefficient storage means for storing the current filter coefficient; and coefficient updating means which adds together said adjustment coefficient and said current filter coefficient from said coefficient storage means to generate the next filter coefficient and provides it to said estimated echo path means and updates said filter coefficient in said coefficient storage means.

Claim 4

The echo canceller of claim 1 or 2 wherein said adjustment matrix is a diagonal matrix having a diagonal component that attenuates at an exponential attenuation ratio γ.

Claim 5

The echo canceller of claim 3 wherein said adjustment matrix is a diagonal matrix having a diagonal component that attenuates in a stepwise manner which approximates an attenuation curve attenuating at said exponential attenuation ratio γ, in steps smaller in number than the number of taps of said estimated echo path means.

Claim 6

The echo canceller of claim 2 wherein, letting said received signal be represented by x(n), said echo by y(n), said adjustment matrix by A, said error signal by e(n), said filter coefficient indicating the impulse response of said estimated echo path means by h(n+1), said adjusted coefficient error covariance matrix by P(n+1) and said gain vector by k(n), said gain vector generating means is a means for calculating ##EQU7## said covariance matrix generating means is a means for calculating, as said updated coefficient error covariance matrix, ν.sup.- [ P(n)-k(n).x(n).sub.T.P(n)]; said adding means is a means for calculating, as said adjusted coefficient error covariance matrix, P(n+1)=ν.sup.-1 [P(n)-k(n).x(n).sub.T.P(n)]+A; said multiplying means calculates the product, e(n)k(n), of said gain vector k(n) and said error signal e(n) as said adjustment coefficient; and said coefficient updating means calculates h(n+1)=h(n)+e(n).k(n), and provides the calculated output, as said filter coefficient, to said estimated echo path means, while at the same time updating the contents of said coefficient storage means, where ##EQU8## α.sub.i =α.sub.0 γ.sup.i-1 for i=1, 2, . . . L, L: number of taps of said digital adaptive filter forming said estimated echo path means, γ: exponential attenuation ratio of the impulse response variation of said echo path, where 0<γ<1, and ν: forgetting factor, where 0<ν≦1.

Claim 7

The echo canceller of claim 1, 2, or 5 wherein said estimated echo path means, said impulse response estimating means and said subtracting means are each provided corresponding to N frequency bands, N being an integer equal to or greater than 2, said echo canceller further including: first sub-band analysis means which divides said received signal into sub-band received signals of said N frequency bands and applies them to said estimated echo path means and said impulse response estimating means corresponding thereto, said estimated echo path means of each sub-band outputting a sub-band echo replica; second sub-band analysis means which divides said echo into sub-band echoes of said N frequency bands and applies them to said subtracting means corresponding thereto, said subtracting means for each frequency band subtracting said sub-band echo replica from the corresponding sub-band echo to generate a sub-band error signal and providing it to said impulse response estimating means for the corresponding frequency band; and sub-band synthesis means which combines said N sub-band error signals into a composite signal and outputs it as said echo-cancelled signal.

Claim 8

An echo cancelling method which repeats steps of: providing a received signal onto an echo path to obtain therefrom an echo; generating an echo replica by applying said received signal to estimated echo path means formed by a digital adaptive filter having an estimated impulse response as a filter coefficient; subtracting said echo replica from said echo and outputting the resulting error signal as an echo-cancelled or echo-free signal; and estimating said filter coefficient of said echo path means as its impulse response on the basis of said error signal and said received signal and iteratively adjusting said estimated filter coefficient by recursive least-squares (RLS) algorithm; wherein said impulse response estimating step includes: a step of storing an adjustment matrix representing an expectation of an impulse response variation of said echo path; a step of adding said adjustment matrix to a coefficient error covariance matrix in said RLS algorithm to generate an adjusted coefficient error covariance matrix; a step of generating a gain vector from said adjusted coefficient error covariance matrix and the vector of said received signal; a step of generating an updated coefficient error covariance matrix from said adjusted coefficient error covariance matrix and said gain vector; a step of multiplying said gain vector and said error signal to generate an adjustment coefficient; a step of storing the current filter coefficient; and a step wherein said adjustment coefficient and said filter coefficient stored in said coefficient storing step are added together to generate the next filter coefficient and said next filter coefficient is provided to said estimated echo path means and is used to update said filter coefficient in said coefficient storage step.

Claim 9

An echo cancelling method which repeats steps of: providing a received signal onto an echo path to obtain therefrom an echo; generating an echo replica by applying said received signal to estimated echo path means formed by a digital adaptive filter having an estimated impulse response as a filter coefficient; subtracting said echo replica from said echo and outputting the resulting error signal as an echo-cancelled or echo-free signal; and estimating said filter coefficient of said echo path means as its impulse response on the basis of said error signal and said received signal and iteratively adjusting said estimated filter coefficient by recursive least-squares (RLS) algorithm; wherein said impulse response estimating step includes: a step of storing an adjustment matrix representing an expectation of an impulse response variation of said echo path; a step of adding said adjustment matrix to a coefficient error covariance matrix in said RLS algorithm to generate an adjusted coefficient error covariance matrix; a step of generating a gain vector from said adjusted coefficient error covariance matrix and the vector of said received signal; a step of generating an updated coefficient error covariance matrix from said adjusted coefficient error covariance matrix and said gain vector; a step of multiplying said gain vector and said error signal to generate an adjustment coefficient; a step of storing the current filter coefficient; and a step wherein said adjustment coefficient and said filter coefficient stored in said coefficient storing step are added together to generate the next filter coefficient and said next filter coefficient is provided to said estimated echo path means and is used to update said filter coefficient in said coefficient storage step.

Claim 10

The echo cancelling method of claim 7 wherein, letting said received signal be represented by x(n), said echo by y(t), said adjustment matrix by A, said error signal by e(n), said filter coefficient expressing the impulse response of said estimated echo path means by h(n+1), said adjusted coefficient error covariance matrix by P(n+1) and said gain vector by k(n), said gain vector generating step is a step of calculating ##EQU9## said covariance matrix generating step is a step of calculating, as said updated coefficient error covariance matrix, ν.sup.- [ P(n)-k(n).x(n).sup.T.P(n)]; said adding means is a step of calculating, as said adjusted coefficient error covariance matrix, P(n+1)=ν.sup.-1 [P(n)-k(n).x(n).sup.T.P(n)]+A; said multiplying step is a step of calculates the product, e(n)k(n), of said gain vector k(n) and said error signal e(n) as said adjustment coefficient; and said coefficient updating step is a step of calculating h(n+1)=h(n)+e(n).k(n), and providing the calculated output, as said filter coefficient, to said estimated echo path means, while at the same time updating said coefficient stored in said coefficient storage step, where ##EQU10## α.sub.i =α.sub.0 γ.sup.i-1 for i=1, 2, . . . , L, L: number of taps of said digital adaptive filter forming said estimated echo path means, γ: exponential attenuation ratio of the impulse response variation of said echo path, where 0<γ<1, and ν: forgetting factor, where 0<ν≦1.