Abstract:

In order to improve the block sparse signal recovery performance, the block restricted isometry constant(RIC)is evaluated by estimating the matrices block-coherence. On this basis, the idea of blocking in union of subspace is introduced into compressive sampling matching pursuit(CoSaMP)in this work, and block CoSaMP(BCoSaMP)is proposed. The support sets were restructured by submatrices, making the restricted isometry property(RIP)relax. It is proved that the larger the block sizes and the fewer the updating block are, the faster the convergence and the bigger the critical sparse ratio are on the conditions of Gaussian measurement matrices. Finally, the recovery simulation of warning radar multi-batches echo is operated, and the results show that a faster speed and higher error stability can be obtained when comparing with other algorithms. Moreover, it has better application value in future.