Skip to Main Content
This paper presents an accurate method to determine the environment model for decentralized detection in sensor networks. We develop a clustering algorithm to classify sensor data and to achieve this model. Then we further enhance the performance of the method in case of noisy sensors, non- identical observations and unreliable communication links. The enhanced algorithm performs an unsupervised hierarchical grouping of the model, threshold detection and outlier removal. The proposed algorithm is verified through simulation.
Date of Conference: 25-28 Sept. 2007