Sparsity Update Subspace Pursuit Algorithm for Compressed Spectrum Sensing

This paper presents an Sparsity Update Subspace Pursuit (SUSP) algorithm for compressed sparse signal reconstruction with unknown sparsity. From practical point of view, the sparsity information is usually unavailable in many applications. In particular, the compressed spectrum sensing application is considered in this paper . The proposed SUSP algorithm begins with sparsity estimation and iteratively updates the sparsity based on the the residual value with subspace pursuit approach. A termination criterion is developed to facilitate the convergence of the sparse update iteration. Moreover, a tail biting rule is devised to refine the reconstruction. Consequently, the sparse signal is recovered and the reconstruction performance is improved. The recovery rate performance is numerically evaluated for both the known sparsity and the unknown sparsity cases. For each case, the recovery mean square errors are also presented for the noisy noisy environment. The results show that the resulting performance outperforms the popular methods using the maximum pursuit or the subspace pursuit based approaches.