Event Detection Method Based on Belief Model for Wireless Sensor Networks

Event area detection is one of the most important applications in complex environment for wireless sensor networks. The nodes in the networks may have faults due to many uncertainty factors( such as temperature,humidity,hardware etc.),which will reduce event detection quality. Considering the spatial correlation for the sensing information of neighboring sensors,a belief model based on network topology was proposed. A Bayesian fault tolerance model was further constructed by synthesizing sensor's own information and its neighbor nodes' information. The model can predict the probability as well as the optimum threshold on whether the event happens. The node makes the final decisions by comparing the two probability values. Simulations show that compared with the optimal threshold decision scheme,the proposed scheme can increase the percentage of the corrected nodes from 65% to 78% when node fault probability is 25%. The scheme can also increase the percentage of the corrected nodes from 50% to 70%when the node fault probability is 30% that reflects better fault-tolerance performance.