Robust Affine Projection Tanh Algorithm and Its Performance Analysis

Abstract

The affine projection algorithm (APA) suffers from performance degradation when dealing with non-Gaussian noises. To address this issue, a novel robust affine projection tanh algorithm (APTA) has been proposed according to a new constructed constraint optimization model with the smoothed and bounded tanh function. Generally, it is difficult to analyze the performance of APTA, since the error vector constrained by the nonlinear function cannot be extracted directly. Therefore, a vector-formed Taylor series expansion is first constructed to decompose the tanh error vector. Then, the steady-state performance, mean and mean-square stability conditions, and transient performance of APTA are carried out for performance analyses. Finally, simulation results verify the effectiveness of theoretical analyses and the superiority of APTA from the aspects of filtering accuracy and robustness.