Variable-length Adaptive Algorithms for Sparse Filters

V. H. Nascimento

DOI: 10.14209/sbrt.2003.598

Evento: XX Simpósio Brasileiro de Telecomunicações (SBrT2003)

Keywords: Sparse adaptive filters convergence rate LMS algorithm

Abstract

Despite their qualities of robustness, low cost, and good tracking performance, for many applications the convergence of the LMS and the normalized LMS algorithms is too slow. Here we analyze a variation of a recently proposed method that speeds up this convergence rate by varying the length of the adaptive filter, taking advantage of the faster convergence rates obtained by short filters. Our results show that variable-length adaptive filters may improve the initial convergence rate of sparse filters when the correlation of the regressor input is not too high. We also provide simulations comparing the new algorithm with NLMS and sparse-signed LMS.

Download