Distributed cooperative spectrum sensing from sub-Nyquist samples for Cognitive Radios

Distributed collaborative spectrum sensing has been considered for Cognitive Radio (CR) in order to cope with fading and shadowing effects that affect a single CR performance, without the communication overhead of centralized cooperation through a fusion center. In this paper, we consider collaborative spectrum sensing by a distributed network of CRs from sub-Nyquist samples to overcome the sampling rate bottleneck of the wideband signals a CR usually deals with. We present a joint reconstruction algorithm, Randomized Distributed Simultaneous Iterative Hard Thresholding (RDSIHT) that adapts the original IHT to block sparse and matrix (simultaneous) inputs, as well as distributed collaboration settings. An observation vector is passed around the network as a random walk process, and updated at each iteration by one of the CRs. Simulations show that our algorithm outperforms a distributed collaborative scheme based on the One-Step Greedy Algorithm (OSGA) using randomized gossip, and that its performance converges to that of its centralized version.