Collaborative Peer-to-Peer Training and Target Classification in Wireless Sensor Networks

Target classification is important in wireless sensor network (WSN). This paper proposes a collaborative peer-to-peer (P2P) training and classifying method with support vector machine (SVM) for WSN. The proposed method incrementally carries out the training process with the collaboration of sensor nodes in P2P paradigm. For decreasing energy consumption and improving accuracy, the collaboration of sensor nodes is implemented by autonomously selecting the proper set of sensor nodes to carry out the training process according to several feasible measures of energy consumption and information utility. Because of the purposeful sensor nodes selection, dynamic collaborative SVM can conquer the inevitable missing rate and false rate of samples in WSN. Results demonstrate that the proposed dynamic collaborative SVM can effectively implement target classification in WSN. It is also verified that the proposed dynamic collaborative SVM has outstanding performance in energy efficiency and time delay.