A new class of gradient-based algorithm by using variable step-size technique

Abstract

This paper presents a new class of gradient-based algorithm for second-order IIR adaptive notch filters (ANF); the algorithm is based on least mean p-power (LMP) error criterion. The variable step-size technique is used to detect high convergence speed and high impulse noise robustness. It is proved here that the proposed algorithm has good properties in both cases, when compared with the previous one (Punchalard et al. 2000). Simulation results are presented to demonstrate the effectiveness of the proposed algorithm.