Decision Fusion Rules in Wireless Sensor Networks Using Fading Channel Statistics

The problem of fusing decisions transmitted over fading and noisy channels in a wireless sensor network is revisited. In a previous work, starting with the assumption of perfect channel knowledge, an optimal likelihood ratio (LR) based fusion rule was developed along with three suboptimum statistics. In this paper, we present a new LR based fusion rule which requires only the knowledge of channel fading statistics instead of the actual channel coefficients. We show that the equal gain combiner (EGC) and the Chair-Varshney fusion rules are approximations of this new rule at low and high channel SNR values, respectively. This new fusion rule has better performance than Chair-Varshney and EGC fusion rules, outperforms the maximum ratio combiner (MRC) at most practical channel SNR values, and is only slightly worse than the optimal LR based fusion rule.