In order to reduce the computational complexity of optimal fusion algorithms with dependent sensor observations

a method for optimizing the system performance based on likelihood ratio decision rules is proposed. The basic idea of the method is to optimize the system performance by constraining the sensor decision rules to be likelihood ratio decision rules

and optimizing the sensor decision thresholds jointly. The necessary condition for the joint optimal sensor decision thresholds is derived for the distributed serial detection systems consisting of N sensors under Bayesian criterion

and the system equations that are satisfied by the optimal sensor decision thresholds are obtained. An iterative algorithm is also proposed to solve for the optimal sensor decision thresholds. Since the joint-optimization of sensor decision rules is simplified to the joint-optimization of decision thresholds

the amount of computational costs is reduced. The algorithm converges well and is more applicable in real applications.