Optimal Decision Fusion when Priori Probabilities and Risk Functions Are Fuzzy

GH Wang,SY Mao,Y He

DOI: https://doi.org/10.1016/j.inffus.2003.03.002

IF: 18.6

2003-01-01

Information Fusion

Abstract:When the priori probabilities and risk functions are fuzzy numbers with trapezoidal membership functions, the optimal decision fusion in the sense of minimum Bayesian risk at the fusion center is considered. Four optimal decision fusion schemes at the fusion center are derived in the sense of minimum Bayesian risk by using the defuzzification methods and the fuzzy ordering criteria. It is discovered that the optimal decision fusion rule in this case is also a weighted sum of local decisions against a threshold, the weights are functions of the probability of miss (or detection) and the probability of false alarm of the local detectors, and the threshold depends not only on the fuzzy a priori probabilities and risk functions but also on the criterion used for defuzzification or ordering fuzzy sets. The Bayesian risk at the fusion center and its membership function are derived. The performance at the fusion center is evaluated with several examples. Finally, some conclusions are drawn.