Fusion of likelihood ratio test based decisions in wireless sensor networks

In this paper, the problem of fusion of decisions transmitted over Rayleigh fading channels in wireless sensor networks (WSNs) is revisited. The likelihood ratio test (LRT) is considered as the optimal fusion rule when applied at the fusion center. However, applying the LRT at the fusion center requires both the channel state information (CSI) and the local sensors performance indices. Acquiring these information is considered as an overhead in a energy and bandwidth constrained systems such as WSNs. To avoid these drawbacks, we propose a modification to the traditional three-layer system model of WSN where the LRT is applied as a local decision method at the sensors level and an equal gain combiner (EGC) is used as a fusion rule at the fusion center. Applying the LRT at the sensor level does not require the CSI or the local sensors performance indices. It only requires the signal-to-noise ratio (SNR). Simulation results show that the performance of the proposed model outperforms the traditional model that applies either the maximal radio combiner (MRC) or Chair-Varshney fusion rules. In addition, it provides comparable performance to the traditional model that applies the LRT at the fusion center.