An on-line MUSIC algorithm with applications to sparse signal reconstruction

Sparse models are a class of systems in which prior knowledge indicates the presence of a majority of zero valued parameters in a model but not the location of the zeros in the parameter vector [1, 2, 3]. Such models occur in a wide range of applications [4, 5, 6, 7, 8, 3]. Many algorithms have been developed for estimating the location of the null parameters. Well known algorithms include, LASSO and Dantzig [9, 10]. These can be thought of as providing a form of regularization. It has been shown that, under appropriate conditions, these algorithms can find the correct location of the null parameters.