ECHO CANCELLING METHOD AND ECHO CANCELLER USING THE SAME

A received signal vector x(n), a coefficient error covariance matrix P(n) from a memory part 12 and a forgetting factor .nu. from a memory part 13 are provided to a gain calculating part 14 to obtain a gain vector k(n). The thus obtained gain vector k(n) and an error e(n) between an echo and an echo replica are multiplied in a multiplying part 16. The multiplied output and a filter coefficient h(n) from a memory part 18 are added together to update the latter. The thus updated filter coefficient is used as the filter coefficient of an estimated echo path (an FIR filter, for example). The coefficient error covariance matrix P(n), the gain vector k(n), the received signal vector x(n) and the forgetting factor .nu. are provided to an updating part 19 to update the coefficient error covariance matrix P(n), and an adjustment matrix A representing an expectation of an impulse response variation of an echo path is added to the updated coefficient error covariance matrix P(n) and the added value is used as a new coefficient error covariance matrix P(n).