| 摘要 |
PURPOSE: To effectively perform echo canceling by finding primary echo, based on plural correlation tap values, predicting an error while narrowing a filter vector, and generating a filter vector updated, based on this predictive error. CONSTITUTION: In the convergence of echo canceling in a digital system, the primary echo in the filter vector is found based on the plural correlation tap values and this primary vector processing area is narrowed to the tap near the primary echo so that the convergence can be increased. Then, a signal vector X(n) 31 is multiplied with vectors 24 and 29 by a filter vector f(n) 32, its scalar product y(n) is subtracted 28 and 29 from a reflected signal s(n) including the echo, and this difference e(n) applies filter error prediction. This predictive e(n) is multiplied 22 and 23 by a basic adaptive rate [α/γx (o) max] and a signal vector x(n) 31, vectors are added 26 and 27 to a vector fn , and an updated filter vector fn+1 is provided.
|
