On the design of soft-decision fusion rule for coding approach in wireless sensor networks

Abstract

In this work, two soft-decision fusion rules, which are respectively named the maximum a priori (MAP) and the suboptimal minimum Euclidean distance (MED) fusion rules, are designed based on a given employed sensor code and associated local classification. Their performance comparison with the distributed classification fusion using soft-decision decoding (DCSD) proposed in an earlier work is also performed. Simulations show that when the number of faulty sensors is small, the MAP fusion rule remains the best at either low sensor observation signal-to-noise ratios (OSNRs) or low communication channel signal-to-noise ratios (CSNRs), and yet, the DCSD fusion rule gives the best performance at middle to high OSNRs and high CSNRs. However, when the number of faulty sensor nodes grows large, the least complex MED fusion rule outperforms the MAP fusion rule at high OSNRs and high CSNRs.