Prediction-Based Algorithm for Event Detection in Wireless Sensor Networks

Event Detection in a special environment is an important application in the wireless sensor networks (WSNs). It´s a hard work due to the sensor nodes are error-prone, and the energy is limited. The existing algorithms work when the events happen, which also assume that those events are spatially correlated. In this paper, we propose a prediction model. As an application of this model, we propose a novel prediction algorithm for the WSNs. We can get an accurate trend measurement for a sensor based on this new algorithm when the event is happening or just happened. In our algorithm, each node transmits its own status which is used to detect and correct the fault nodes to every neighbor. Our experimental results show that the algorithm can detect above 98% of the fault nodes and 92% area of the event region when 20% nodes are fault. Moreover, our algorithm is more energy efficient since it can get rid of frequent measurements exchange.