A Normalized Filtered-x Generalized Fractional Lower Order Moment Adaptive Algorithm for Impulsive ANC Systems

Abstract

This paper proposes an efficient algorithm for impulsive active noise control (IANC) systems. The impulsive sources cannot be modeled by Gaussian distribution, and hence the standard adaptive algorithm based on second order statistics would give poor performance or even fail to converge. One solution is to derive adaptive algorithm by minimizing a fractional low order moment, resulting in the famous filtered-x least mean p-power (FxLMP) algorithm. The proposed algorithm discussed in this paper is based on a previously proposed generalized FxLMP algorithm. The key idea here is to introduce a variable step-size using a convex-combination approach. A large value is used at the start-up of IANC system to achieve a fast convergence speed. As the AINC system converges, the step-size automatically reduces to a small value to improve the steady-state noise reduction performance. Simulations demonstrate the effectiveness of the proposed algorithm.