New sub-band proportionate forward adaptive algorithm for noise reduction in acoustical dispersive-and-sparse environments

Abstract

In this study, we address the problem of acoustic noise reduction by two-channel forward adaptive filtering algorithm. In noise reduction applications, the normalized least mean square (NLMS) algorithm is recently implemented on the two-channel blind source separation (BSS) technique. Exactly in this paper, we propose a new implementation of two-channel forward BSS technique based on proportionate adaptive filtering algorithm in sub-band form. This new implementation is proposed exactly to resolve the problem of proportionate forward NLMS algorithm in term of convergence rate and speech quality when the acoustic environment is characterized by dispersive or sparse impulse responses. To prove the good performance of our proposed sub-band proportionate Forward NLMS algorithm (SP-FNLMS), we compare its performances with the full-band proportionate and non-proportionate versions exactly in four acoustical environments, more dispersive, dispersive, sparse and finally more sparse.