Noise Reduction in Multi-Channel Feed Forward ANC System with Online Secondary Path Modeling

Abstract

Adaptive algorithms have been used, since long, for low frequency active noise control (ANC) applications. However, their main area of application has been Gaussian noise. In recent past, some work has been done on impulsive noise (IN) cancellation in single channel ANC systems. This inscription deals with IN in multichannel (MC) ANC Systems with Online Secondary Path Modelling (OSPM) employing adaptive algorithms for the first time. It compares performance of various existing techniques belonging to varied computational complexity range and proposes four new methods, namely: Filtered-x Recursive Least Square-Variable Step Size Least Mean Square (FxRLS-VSSLMS), VSSLMS-VSSLMS, Filtered-x Least Mean Absolute Third-VSSLMS (FxLMAT-VSSLMS) and Normalized Step size Modified FxLMAT-VSSLMS (NSS MFxLMAT-VSSLMS) to deal with modest to very high IN. First proposed FxRLS-VSSLMS method successfully alleviated IN but with higher computational complexity. To reduce complexity, second proposed VSSLMS-VSSLMS method was utilized which achieved same steady state error as of proposed FxRLS-VSSLMS but convergence speed was effected. In order to balance computational complexity and convergence speed, third proposed FxLMAT-VSSLMS method is employed against IN which although achieved same steady state error but suffered with instability and creeping convergence speed. Lastly, a modified version of FxLMAT is employed in proposed NSS MFxLMAT-VSSLMS method which not only achieved same steady state error but also exhibited stability and robustness against IN with faster convergence and least computational complexity. Results of simulation proved that these proposed methods demonstrated better performance in terms of fast convergence speed, lowest steady state error, robustness and stability under impulsive environment in addition to modeling accuracy for stationary as well as non-stationary environment besides reducing computational complexity many folds.