Decentralized Neyman-Pearson Test with Belief Propagation for Peer-to-Peer Collaborative Spectrum Sensing

In this paper we propose a decentralized approach for cooperative signal detection, based on peer-to-peer collaboration among sensor nodes. The proposed method combines belief propagation, implemented in a distributed fashion through the exchange of local messages to and from neighboring nodes, with a Neyman-Pearson framework, that allows control over the false-alarm rate of each node. At the same time, nodes gradually learn their degree of correlation with neighbors, and clusters of nodes under homogeneous conditions are formed automatically. The performance of the resulting va Neyman-Pearson belief propagation" (NP-BP) algorithm is shown to be nearly equivalent to that of cooperative energy detection applied separately at each cluster. Thanks to its decentralized structure, NP-BP provides improved robustness, flexibility, and scalability compared to traditional, centralized schemes. In addition, its ability to adaptively form clusters makes the algorithm suitable for heterogeneous or time-varying radio environments.