LAUSR

Abstract The communication in teleconferencing systems is influenced by acoustic echo and noise signals. Several techniques have been proposed, such as the blind source separation (BSS) that has recently been integrated with adaptive filtering algorithms to reduce the acoustic noise. This combined technique can be used to extract the speech signal from only the mixed noisy signals alone. In this present study, we propose a new extended two-sensors noise reduction algorithm based on a decorrelation criterion implemented on the recursive BSS technique in sub-band form. In this version, we propose to use separated adaptive sub-filters that are updated by the minimization of the cross-correlation between output sub-signals. Simulation results of the proposed algorithms confirm its better performances compared to full-band versions in term of convergence rate based on Mean Square Error (MSE).