Improved filtered-s least mean square/fourth algorithm for nonlinear active noise control system

Abstract

This paper presents a novel approach for nonlinear active noise control (ANC) systems that addresses the limitations of traditional linear ANC systems, which fail to consider nonlinear factors. A nonlinear ANC system is proposed for non-stationary noise with a nonlinear secondary path using the improved filtered-s least mean square/fourth (IFSLMS/F) algorithm. This system utilizes the real-time complementary ensemble empirical modal decomposition (CEEMD) for noise decomposition, trigonometric functional expansions of the functional link artificial neural network (FLANN) filter, and adaptive weight updating. The IFSLMS/F algorithm uses the fast convergence of the variable step size algorithm to efficiently reduce error perturbation by performing power normalization on the error signal. The performance and robustness of the system have been validated by simulations and experiments.