A multi-sensor decision fusion scheme is presented in which the probabilities associated with the local sensor decisions are known to vary in a nonrandom fashion around their design values. The uncertainties associated with the local decisions are modeled by means of fuzzy sets. A Bayesian approach is used to design the optimum fusion rule for the case where the local sensor decisions are statistically independent across the sensors. In order to reach a crisp decision, the global Bayesian risk is defuzzified using a criterion for mapping fuzzy sets on to the real line. The performance of the optimum fusion rule obtained is illustrated by means of a numerical example.

In this paper, we present a new upper bound on the minimum probability of error of Bayesian decision systems for statistical pattern recognition. This new bound is continuous everywhere and is shown to be tighter than several existing bounds such as the Bhattacharyya and the Bayesian bounds. Numerical results are also presented.>