Fast System Identification Using Affine Projection and a Critically Sampled Subband Adaptive Filter

Abstract

Classical least-mean-square (LMS) adaptive filtering algorithms for system identification are popular and conceptually simple. In many applications subband adaptive filter structures have been shown to be superior computationally and performance-wise. This paper presents a novel subband affine projection algorithm (APA) suitable for use within a recently proposed adaptive filter structure employing critically sampled filter banks. The algorithm is described in the context of measuring a room impulse response for acoustic echo cancellation in hands-free telephony. Experimental results with speech input signals in a conference room show that a four-channel subband adaptive filter with subband APA can achieve an average 5 dB lower mean square error than a subband normalized LMS