Wireless sensor network faulty scenes diagnosis using high dimensional Neighborhood Hidden Conditional Random Field

Wireless sensor networks are widely deployed in different industrial applications. Fault diagnosis of sensors is essential for maintaining a robust WSN operation. In this paper, we show faulty sensors diagnosis can be transformed into a pattern classification problem. We also introduce an efficient algorithm, called, Neighborhood Hidden Conditional Random Field, to recognize sensor states and the faulty scene of an WSN. Compared to conventional methods, the proposed diagnosis method incorporating hidden states can estimate the posterior probability of different faulty scenes. In addition, nearest neighbors are selected for estimating the dependencies among sensors, and the dependencies are subsequently used for diagnosing faulty scenes. Simulations based on an WSN are presented to show the effectiveness of the proposed methods. The results demonstrate that our proposed algorithm can achieve better classification performance compared with other state-of-art methods.