Multiple hypotheses testing method for distributed multisensor systems

In this paper, we propose a two-layer sensor fusion scheme for multiple hyp otheses multisensor systems. To reflect reality in decision making, uncerta in decision regions are introduced in the hypotheses testing process. The e ntire decision space is partitioned into distinct regions of "correct", "un certain" and "incorrect" regions. The first layer of decision is made by ea ch sensor indepedently based on a set of optimal decision rules. The fusion process is performed by treating the fusion center as an additional "virtu al" sensor to the system. This "virtual" sensor makes decision based on the decisions reached by the set of sensors in the system. The optimal decisio n rules are derived by minimizing the Bayes risk function. As a consequence , the performance of the system as well as individual sensors can be quanti fied by the probabilities of correct, incorrect and uncertain decisions. Nu merical examples of three hypotheses, two and four sensor systems are prese nted to illustrate the proposed scheme.