Block backtracking-based matching pursuit for arbitrary block sparse signal recovery

Compressed sensing can exactly reconstruct sparse or compressible signals. In practice, many sparse signals have block sparse structures. To recover block sparse signals, a block backtracking-based matching pursuit algorithm is proposed, which chooses possible correct atoms by using a block processing and deletes unreliable atoms by a backtracking mechanism. The proposed algorithm doesn't require the prior knowledge about the sparse structure of signal. Experiments demonstrate that the proposed algorithm outperforms some other greedy recovery methods.