A robust double-channel affine projection sign algorithm for blind acoustic interferences cancellation

Abstract

This brief proposes a double-channel direct affine projection sign (DC-DAPS) algorithm for blind speech enhancement applications, especially in the presence of strong acoustic noise, and intensive non-Gaussian impulsive interferences. We also derive a full analysis of the stability of the proposed DC-DAPS algorithm. The performance evaluation of the proposed algorithm confirms its accuracy even under different noisy scenarios, and demonstrates that our approach outperforms the double-channel direct affine projection (DC-DAP), and the double-channel direct normalized least mean square (DC-DNLMS) algorithms. Simulation results showed that the proposed DC-DAPS algorithm could effectively achieve improved performance in combating acoustic noise and impulsive interferences, and speeding up the convergence rate even with highly correlated input signals such as the speech signal.