A Novel Decision Fusion Scheme with Feedback in Neyman-Pearson Detection Systems

Abstract

Information fusion brings great advantages in multi-sensor detection systems and has attracted much attention. Regarding decision level fusion, in many existing literatures, the fusion center (FC) gives a global decision via a likelihood ratio (LR) test where the LR function is compared with a constant threshold. Most of these can be viewed as one-stage decision fusion schemes because the FC does not utilize the historical information in a continuous period of time. In this paper, we propose a novel multi-stage decision fusion scheme with feedback from the view of the Neyman-Pearson (N-P) criterion. In the proposed scheme, at each stage, the FC selects one threshold from two alternative values based on the feedback of the previous stage’s global decision to perform the LR test. Then we prove the convergence of the global detection probability and the false alarm probability when the true state of the target remains unchanged. For the decision fusion of two homogeneous sensors, we derive the optimal alternative thresholds under the N-P criterion. Simulation results show that the proposed scheme can effectively improve the performance of target detection.