A Improved Gradient Projection Mrthod For Signal Reconstruction

Gradient projection (GP) algorithm has already been used for signal recon-struction. Firstly§we introduce GPSR-BB gradient projection approach§it trans-forms the l1-minimization problem to the bound-constrained quadratic program-ming(BCQP) for signal reconstruction. We propose a new gradient projection al-gorithm for signal reconstruction by changing the line search conditions of GPSR-BB approach. Then we prove that the algorithm is well defined and the conver-gence properties of the algorithm are presented: algorithm will be terminated ata solution of BCQP or an arbitrary limit point of the iterative sequence generatedby this algorithm converges to a solution of BCQP. Numerical experiments areconducted with the method. We show that the new algorithm get better numericalexperiments results in CPU time and number of iterations.