Evaluation of Performance Bounds in Distributed Detection.

A generalized version of the majority voting rule is proposed to speed up the calculation of global error probabilities in parallel sensor networks, which otherwise requires exponentially many computations in general. The motivation behind proposing such a fusion rule is to inform the other researchers that the recently available performance bounds are not tight as claimed and can only be used for very limited decentralized detection problems. The proposed fusion rule can be computed in quasi-linear time for identically and quasi-quadratic time for non-identically distributed independent sensor observations. Effectiveness of the proposed fusion rule is illustrated with comparative experiments for identical and non-identical sensors over varying size of networks and various quantization levels.