Distributed decision fusion using the Neyman-Pearson criterion

Performance of a parallel binary decision fusion architecture is considered where both the local detectors and the Decision Fusion Center use the Neyman-Pearson criterion. The architecture comprises N local detectors in parallel, each sending a binary decision to a fusion center for integration. The algorithm designed here fixes the global false alarm rate and attempts to compute the local detector thresholds and the global fusion rule that achieve the maximum global detection probability. The key computational requirement is to find the roots of a certain Nth order polynomial. We compare the performance of our method with the performance of the Person-by-Person Optimization (PBPO) approach and that of a centralized detection scheme.