Decision Fusion In Sensor Networks Research Articles

Belief Function Based Decision Fusion for Decentralized Target Classification in Wireless Sensor Networks.

Decision fusion in sensor networks enables sensors to improve classification accuracy while reducing the energy consumption and bandwidth demand for data transmission. In this paper, we focus on the decentralized multi-class classification fusion problem in wireless sensor networks (WSNs) and a new simple but effective decision fusion rule based on belief function theory is proposed. Unlike existing belief function based decision fusion schemes, the proposed approach is compatible with any type of classifier because the basic belief assignments (BBAs) of each sensor are constructed on the basis of the classifier’s training output confusion matrix and real-time observations. We also derive explicit global BBA in the fusion center under Dempster’s combinational rule, making the decision making operation in the fusion center greatly simplified. Also, sending the whole BBA structure to the fusion center is avoided. Experimental results demonstrate that the proposed fusion rule has better performance in fusion accuracy compared with the naïve Bayes rule and weighted majority voting rule.

Locating hot nodes and data routing for efficient decision fusion in sensor networks

This paper presents a fundamentally new approach to integrating local decisions from various nodes and efficiently routing data in sensor networks. By classifying the nodes in the sensor field as “hot” or “cold” in accordance with whether or not they sense the target, we are able to concentrate on a smaller set of nodes and gear the routing of data to and from the sink to a fraction of the nodes that exist in the network. The introduction of this intermediary step is fundamentally new and allows for efficient and meaningful fusion and routing. This is made possible through the use of a novel Markov Random Field (MRF) approach, which, to the best of our knowledge, has never been applied to sensor networks, in combination with Maximum A Posteriori Probability (MAP) stochastic relaxation tools to flag out the “hot” nodes in the network, and to optimally combine their data and decisions towards an integrated and collaborative global decision fusion. This global decision supersedes all local decisions, and provides the basis for efficient use of the sensed data. Because of the MRF local nature, nodes need not see or interact with other nodes in the sensor network beyond their immediate neighborhood, which can either be defined in terms of distance between nodes or communication connectivity, hence adding to the flexibility of dealing with irregular and varying sensor topologies, and also minimizing node power usage and providing for easy scalability. The routing of the “hot” nodes’ data is confined to a cone of nodes and power constraints are taken into account. We also use the found location of the centroid of the hot nodes over time to track the movement of the target(s). This is achieved by using the segmentation at time t as an initial state in the stochastic MAP relaxation at time t + Δ t.