Stabilized Stepwise Orthogonal Matching Pursuit for Sparse Signal Approximation
Mingjiang Wang,Guanghong Liu,De Zhang,Kuoye Han,Yanmin Chen
DOI: https://doi.org/10.1088/1742-6596/910/1/012038
2017-01-01
Abstract:Orthogonal Matching Pursuit (OMP) algorithm is equipped with the capability to decompose any signal into a linear expansion of waveforms, which are selected from a redundant functional dictionary. Nevertheless, classical OMP algorithm suffers a heavy computational burden due to its single element selection strategy in each repetition. Recently, an accelerated implementation called stage wise orthogonal matching pursuit (StOMP) algorithm has been proposed through exploiting a multiple elements selection scheme based on an iterative threshold. However, as the defined threshold is a function of an empirical and undetermined parameter, such a reconstruction scheme is not optimal and the algorithm may get obstructed in some specific conditions. This manuscript presents an adaptive threshold selection strategy which takes signal structure into consideration and furthermore, a regularized iterative framework for sparse signal approximation is suggested. Compared with classical StOMP approach, these efforts can provide robust and more attractive approximation performance for sparse signal recoveries. Experimental results present the substantial improvements of these optimizations.