New Improved Algorithms for Compressive Sensing Based on $\ell_{p}$ Norm

摘要: A new algorithm for the reconstruction of sparse signals, which is referred to as lp-regularized least squares ( lp-RLS) algorithm, proposed. The based on minimization a smoothed lp-norm regularized square error with p <; 1 . It uses conjugate-gradient (CG) optimization method in sequential strategy that involves two-parameter continuation technique. An improved version also proposed, entails bisection technique optimizes an inherent regularization parameter. Extensive simulation results show offers signal performance and requires reduced computational effort relative several state-of-the-art competing algorithms. lp-RLS better than basic version, although this achieved at cost increased effort.