Improved Filtered-x Least Mean Kurtosis Algorithm for Active Noise Control

The least mean kurtosis (LMK) algorithm has been successfully applied to linear system identification. It outperforms the conventional least mean square method for Gaussian and non-Gaussian noises. The purpose of this work is to apply the LMK algorithm to active noise control (ANC) systems, i.e., to develop a filtered-x LMK (FxLMK) algorithm. The proposed FxLMK algorithm is robust against the attenuation of various noise interferences in ANC system. To further improve the noise reduction performance, a recursive sampled variance method is incorporated into the FxLMK algorithm, resulting in an improved FxLMK (IFxLMK) algorithm without significantly increasing the computational burden. Simulations in the various noise inputs show the effectiveness of the proposed algorithms. Moreover, the traction substation noise control problem is considered in our simulations, and an overall system scheme and simulation studies are conducted. Simulation results demonstrate that the IFxLMK algorithm can improve the convergence rate of adaptive filter and improve the performance of noise reduction.