Particle swarm optimization based novel adaptive step-size FxLMS algorithm with reference signal smoothing processor for feedforward active noise control systems

Abstract

The unimproved filtered-x least mean square (FxLMS) algorithm is not enough to guarantee the convergence rate and stability when outliers appear in the reference signal Xn. To make up for this deficiency, a novel adaptive step-size FxLMS with reference signal smoothing processor (NASFSxLMS) algorithm is proposed to accelerate the convergence rate and weaken the impact of outliers. The reference signal smoothing processor integrates the Moving-Median filter and the Hampel filter to preprocess different types of reference signals. The novel adaptive step-size is developed by normalizing the step-size adjustment factor which takes the Euclidean Norm of Xn and error signal en into consideration. What’s more, different coefficients of the step-size adjustment factor h(n) in the normalization function result in different noise reduction effects. To further ameliorate the noise reduction performance of the proposed algorithm, we combine the NASFSxLMS algorithm with the particle swarm optimization (NASFSxLMS-PSO) algorithm to get the optimal coefficients of h(n). Simulation results demonstrate the faster convergence rate, enhanced stability and higher noise reduction of the proposed algorithms under different types of reference signals.