Sparse component analysis for linear mixed models

M. HURTADO; N. VON ELLENRIEDER; C. MURAVCHIK; A. NEHORAI

Workshop; IEEE Sensor Array and Multichannel Signal Processing Workshop; 2010

When seeking for a sparse solution of a linearmodel, a commontechnique is the search of a solution with minimumL1 norm. In this paper, we present a new approach for thecase of sparse linear mixed models. We combine the EMalgorithm for solving the inverse problem with a decisiontest that guarantees sparseness by eliminating the statisticallynull components of the solution. We address its performanceby means of simulations and illustrate its use withreal radar data demonstrating its potential applications.