Multichannel Feed-Forward Active Noise Cancellation Systems

Abstract

Multichannel Active noise cancellation (MC-ANC) based on the FxLMS method is extensively used to mitigate noise in multiple sensor systems. However, FxLMS based MC-ANC system face divergence and stability issues in the presence of impulsive noise (IN). Motivated by the excellent performance of filtered x least mean absolute third FxLMAT method in the single channel (SC) ANC system of IN, we tested it in MCFxLMAT. Contrary to the previously achieved results with FxLMAT in SC-ANC of IN, FxLMAT didn’t perform well in MC-ANC. Therefore, three variants of FxLMAT are proposed in this paper. The first alteration, i.e., Proposed MC filtered x normalized LMAT (MC-FxNLMAT), is based on step size normalization with respect to the input signal as it exhibited poor convergence and robustness in the presence of IN environment. Hence, the second modification, i.e., MC filtered x robust normalized LMAT (MC-FxRNLMAT), is suggested to improve the convergence of MC-FxNLMAT by normalizing with the error signal but still lacks robustness. To overcome the problems of prior proposed methods third variant, i.e., MC-Binormalized FxLMAT is devised, which normalizes step size with respect to both input and error signal. The proposed MC- Binormalized FxLMAT algorithm depicts better stability and fast convergence at the cost of high complexity as compared to the tested method for stationary and nonstationary in rigorous simulations.