Energy-Efficient Pattern Recognition Approach for Wireless Sensor Networks

Event detection and monitoring is an important application class for wireless sensor networks. Traditionally, sensory data are collected and processed at the base-station. Conveying large amounts of multidimensional sensory data is however impractical in resource-constrained sensor networks. In this paper we propose to convert event detection into pattern recognition that is particularly suited for sensor networks. Individual sensory measurements of sensor nodes are integrated into high-level event pattern, and used for recovering the state of the monitored environment. The pattern storage and pattern recognition operations are performed in a distributed manner within the network. Furthermore, a sleep mode strategy is incorporated for improving performance and prolonging the lifetime of the sensor network.