Estimación de parámetros en redes inalámbricas de detección distribuida

Abstract

This work presents an algorithm to blindly estimate the model parameters of decision fusion systems over wireless sensor networks. In particular, it considers the so-called canonical distributed detection systems, where the sensors report their decisions to the fusion center (FC), through independent binary symmetric channels. Then, the FC makes the final decision by combining the noisy sensor decisions according to a certain fusion rule. We consider fully heterogeneous networks where the sensors can have different probabilities of detection and false-alarm, and the reporting channels can have different crossover probabilities. When the FC knows all these model parameters the optimal fusion rule is the likelihood ratio (LR) test. But the likelihood ratio depends on the model parameters, which may be unknown (all or some of them) in many practical cases, making the LR test inapplicable. In this work, we present an algorithm for the FC to blindly learn the sensor probabilities of detection from the noisy sensor decisions received after a number of sensing periods. The algorithm can also estimate the prior probabilities of the null and alternative hypothesis when they are unknown by the FC. Then, based on the estimates of these model parameters, a channel-aware fusion rule is derived. Simulation results show that, after sufficient sensing periods, the model parameter estimates are accurate enough for the fusion rule to exhibit near-optimal detection performance