Block filtered-s least mean square algorithm for active control of non-linear noise systems

Abstract

In this study, the authors propose a block formulation of an algorithm, called the block-filtered-s LMS (BFSLMS) algorithm, for active control of non-linear noise processes for a multichannel setup. A reduced structure of the fast Fourier transform (FFT)-based BFSLMS-M (FBFSLMS-M) algorithm has also been studied. From this, the multichannel block filtered-x LMS (FBFXLMS-M) algorithm has been derived as a special case. The simulation results show that these algorithms have a matching performance with the already existing algorithms, with a relatively low computational complexity. A reduced structure delayless FBFSLMS algorithm for the multichannel case has also been developed which has a lesser computational complexity than its time-domain counterpart. Apart from this, it has no delay which, in general, is inherent in the block adaptive algorithms.