Parametrização Automática do Algoritmo l0-LMS com Relação à Razão Sinal-Ruído

Jéssica Bartholdy Sanson, Mariane Rembold Petraglia, Diego Barreto Haddad

DOI: 10.14209/sbrt.2015.150

Evento: XXXIII Simpósio Brasileiro de Telecomunicações (SBrT2015)

Keywords: Norm Least Mean Square (LMS) Adaptive Filtering Sparsity

Abstract

The l0-LMS is an adaptive algorithms recently proposed for the identification of sparse systems. Its strategy consists in modifying the cost function of the standard LMS algorithm by adding a term that penalizes non-sparse solutions. Compared to its precursors, this algorithm has proven very competitive. However, the advantages of the l0-LMS algorithm usually depend upon a judicious selection of its parameters, thereby avoiding performance evenworse than the traditional LMS. This paper proposes, in the system identification context, a procedure for automatic adjustment (dependent on the signal- to-noise ratio) of the parameter related to the penalizing term, in order to ensure an advantageous performance of the algorithm l0-LMS

Download