Distributed M-ary hypothesis testing for decision fusion in multiple-input multiple-output wireless sensor networks

In this study, the authors study binary decision fusion over a shared Rayleigh fading channel with multiple antennas at the decision fusion centre (DFC) in wireless sensor networks. Three fusion rules are derived for the DFC in the case of distributed M-ary hypothesis testing, where M is the number of hypothesis to be classified. Namely, the optimum maximum a posteriori (MAP) rule, the augmented quadratic discriminant analysis (A-QDA) rule and MAP observation bound. A comparative simulation study is carried out between the proposed fusion rules in-terms of detection performance and receiver operating characteristic (ROC) curves, where several parameters are taken into account such as the number of antennas, number of local detectors, number of hypothesis and signal-to-noise ratio. Simulation results show that the optimum (MAP) rule has better detection performance than A-QDA rule. In addition, increasing the number of antennas will improve the detection performance up to a saturation level, while increasing the number of the hypothesis will deteriorate the detection performance.