Intelligent protocols based on sensor signal change detection

Embedded into the communication protocols, the signal change detection will allow data compression for improving network efficiency. It enhances reliability and security also. The proposed change detector employs a neural network function prediction in order to determine if the sensor outputs have changed In addition to the change detection system, a modification to a standard neural network function predictor is proposed that allows the change detection system to quickly learn how to accurately predict next sensor outputs. The parameter choice and the relationship between the threshold values and false alarm and change missing rates are studied The protocol is implemented and tested in real life environments with sensor networks built from Crossbow MICA-2 motes. Sensor network change detection system, which is designed to become a protocol core utility, is described The test results are analyzed and recommendations on applications are derived.