From Centralized Data Fusion to Hard- and Soft-Decision Fusion in Multisensor Distributed Detection Systems a New Multiple-Bit Decision Fusion Approach

We consider multisensor distributed detection systems with data fusion which consist of a number of distributed local sensors and a data fusion center. The data fusion center receives information from the local sensors and fuses them to obtain a final global decision on one of two hypotheses. In this paper, we first revisit and formulate the three major approaches of data fusion in multisensor distributed detection systems in the general case of $n$ distributed sensors; namely, centralized data fusion, hard-decision fusion, and soft-decision fusion. The formulation is based on the Neyman-Pearson criterion assuming that the local sensors are independent and observe the same phenomenon. Then, we propose a new multiple-bit decision fusion approach in multisensor distributed detection systems with data fusion. Unlike most of the published studies related to hard-decision fusion, the proposed approach is based on fusing multi-bit decisions instead of fusing one-bit decisions. Unlike most of the published studies related to soft-decision fusion, the proposed approach can be applied easily to identical and non-identical sensors and can be extended easily to any number of sensors and any number of bits per decision. Furthermore, it does not depend on the derivatives of the equations related the false alarm and detection probabilities of the individual sensors and does not require excessive computations as the number of bits per decisions increases. The proposed multiple-bit decision fusion approach is derived and its global performance is evaluated and compared to other fusion approaches in case of noisy Gaussian and Rayleigh distributed observations.