Gradually Atom Pruning for Sparse Reconstruction and Extension to Correlated Sparsity

We propose a new algorithm for recovery of sparse signals from their compressively sensed samples. The proposed algorithm benefits from the strategy of gradual movement toestimate the positions of non-zero samples of sparse signal. We decompose each sample of signal into two variables, namelyvalue and detector, by a weighted exponential function. We update these new variables using gradient descent method. Like the traditional compressed sensing algorithms, the first variableis used to solve the Least Absolute Shrinkage and Selection Operator (Lasso) problem. As a new strategy, the second variableparticipates in the regularization term of the Lasso (l1 norm) that gradually detects the non-zero elements. The presence of thesecond variable enables us to extend the corresponding vector of the first variable to matrix form. This makes possible use of thecorrelation matrix for a heuristic search in the case that there are correlations among the samples of signal. We compare the performance of the new algorithm with various algorithms foruncorrelated and correlated sparsity. The results indicate the efficiency of the proposed methods

Keywords:

Smoothed l0 , l1 minimization , compressed sensing , reconstruction algorithm , correlated sparsity