Distributed compressed sensing for block-sparse signals

To address the problems of high sampling rates, shadow fading and additive noise from the receiver, in this paper, a distributed compressed sampling (DCS) and centralized reconstruction approach which utilize the spatial diversity against fading channels is proposed. Unlike traditional centralized reconstruction, in this paper, we centralized recover the spectrums by exploiting the block-sparsity which is rather prevalent in multi-band signals. In DCS, first each CR samples the signals with a sub-Nyquist sampling rate independently, then the sampled data are uploaded to the fusion center (FC), finally FC reconstructs these data simultaneously. To exploit the block-sparsity, two new centralized recovery algorithms simultaneous block orthogonal matching pursuit (S-BOMP) and simultaneous binary tree based block orthogonal matching pursuit (S-BTBOMP) are developed. Simulation results show they outperform existing simultaneous recovery algorithms which don´t take block-sparsity into consideration.