Decision fusion rules for multiple hypotheses in heterogeneous wireless sensor networks

Abstract

In this paper, a new decision fusion algorithm based on Dempster-Shafer theory (DST) and Bayes criterion for multiple hypotheses in heterogeneous wireless sensor network is proposed. Firstly, the two layer sensor fusion scheme is put forward to study the decision fusion rules of the multiple hypotheses multisensory systems. In the first layer, the optimal local decision fusion rule and the suboptimal decision fusion rule named expect of mean-square error (EMSE) for single channel where a flat fading channel is considered are deduced by minimizing the Bayes risk. Then, the DST is applied in the global decision fusion based on the local decision results. Meanwhile, a model for the basic probability assignment (BPA) of Dempster-Shafer (D-S) evidence theory is built, which not only updates BPAs without depending on expert systems, but also makes the center fusion more intelligent. Finally, the new improved Dempster-Shafer (D-S) fusion algorithm based on similarity coefficient weighting (DSSC) is proposed when the problem of conflict evidence is considered. Simulation shows that the new decision fusion algorithm provides much better detection performance than those of the K out of N (KN) and the original D-S decision fusion rule.