Abstract:

Compressed sensing is a new way of signal sampling and data compression.The subspace pursuit algorithm has higher efficiency and precision in the compressed sensing reconstruction algorithms,but it needs the sparsity of the signal as a priori information.And if the sparsity estimation is not accurate enough,it will reduce the algorithm reconstruction effect.Aiming at this problem,a two stagewise weak selection-based subspace pursuit (TSWSP) algorithm was proposed,which didn’t need to know the sparsity of the signal in advance.The first weak selection adaptively selected the initial atom candidate set,and the second weak selection adaptively culled the wrong atoms that may had been previously selected from the current atom support set,and finally it selected a plurality of related atoms from the current atom candidate set to join the atom support set by the backtracking method.Simulation analysis shows that the proposed algorithm can reconstruct one-dimensional random signals and two-dimensional image signals accurately with unknown sparsity,and it has high stability,compared with OMP,SWOMP,BAOMP,SAMP and SP algorithm,the mean-square erroris reduced by 60.5% to 99.1%,the peak signal-to-noise ratio is improved by 2.1% to 34.3%.

Key words: compressed sensing, subspace pursuit, stagewise weak selection, signal reconstruction