Blind sparsity weak subspace pursuit for compressed sensing

Abstract

A type of back-track algorithm has recently been proposed to handle sparse reconstruction problems arising in compressed sensing. This sort of algorithm is attractive owing to its recovery performance; however, it requires the sparsity level to be known a priori. A blind sparsity algorithm is presented based on subspace pursuit and weak pruning, and its convergence is demonstrated. The experiments validate the proposed algorithm's superior performance to that of several other back-tracking-type and optimisation methods. © The Institution of Engineering and Technology 2013.