Energy-efficient k-class coverage for collaborative classification in Wireless Audio Sensor Networks

Object/event classification is an important aspect in the applications of Wireless Audio Sensor Networks (WASNs). In order to reduce the computational burdens of individual sensors, distributed classification method is adopted, i.e., the individual sensors only perform binary classification and send out a binary decision to the fusion center. However, it brings about a new coverage problem, namely, how to deploy sensors so that all locations in the surveillance region are covered by multiple types of sensors with different sensing radii for collaborative classification. We call it k-class coverage problem. In this paper, we propose two algorithms: straight-forward algorithm and divide-and-conquer algorithm. The first algorithm converts the problem into k-class set cover problem, and the second algorithm converts the problem into solving set 1-cover problem k times. Greedy strategies are used in both of two algorithms. Performance analysis and simulation results are provided for evaluating two algorithms.