Power allocation for decision fusion in wireless sensor networks by the Cauchy-Schwartz divergence

The statistical distances or similarity measures are fundamental tools for solving a wide range of statistical signal processing problems. In this paper, we consider a novel information theoretic divergence as a performance criterion to optimize decision fusion over a wireless sensor network. Specifically, the Cauchy-Schwartz divergence between probability densities of the received signal under different hypotheses is used. This measure can lead to an analytic closed form expression for a mixture of Gaussians, while most of the well-known divergences cannot. Both orthogonal and nonorthogonal communication channels are considered. Simulation results validate the theoretically claimed improvement in the performance.