On the detection of sparse signals with sensor networks based on subspace pursuit

In this paper, we consider the problem of distributed detection of sparse signals with a sensor network. Due to practical constraints on communication bandwidth and computational capacity, detection of sparse signals in a distributed manner is more efficient than centralized processing in terms of communication and computation. We develop a greedy algorithm named distributed subspace pursuit (DSP) for distributed detection of sparse signals. In the proposed approach, each node computes an estimate for the sparse support iteratively using the subspace pursuit (SP) algorithm and transmits a condensed message to a fusion center during each iteration to compute a decision statistic. Experimental results show that reliable detection of sparse signals can be obtained by the proposed approach with a very small number of iterations per node. We further demonstrate the superiority of our algorithms to the most related sparsity-driven methods.