ADAPTIVE SUB-BAND FILTERING

摘要

A digital signal adaptive filtering method and device are described. The transmitted signal (xn), which is representative of a post-transmission distorted source signal (dn), is adaptively filtered (W*) (1001) to generate a filtered transmitted digital signal (yn). An error signal (en) is formed (1002), which is representative of the algebraic difference between the source signal (dn) and the filtered transmitted digital signal (yn), and which is subdivided (1003), through an orthogonal breaking-up step resulting in a number (N) of frequency sub-bands, into sub-band error signals (ei,n), i ELEMENT [0,N-1]. Each sub-band error signal is sub-sampled in a ratio of (N) and the sub-sampled sub-band error signals (eik) are weighted (1004) with coefficients ( lambda i). The transmitted digital signal (xn) is subdivided (1005) into (N) sub-bands (xi,n) and an adaptation function ( DELTA W*) of the adaptive filter is established (1007) as the sum of the product of each conjugated sub-band component of the transmitted digital signal and the sub-sampled weighted sub-band error signal, to which an adaptive pitch ( mu ) has been allocated. The invention is useful for speech signal echo equalising and cancelling.