Performance Evaluation of Decision Fusion in Wireless Sensor Networks
- 1 March 2006
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 69-74
Abstract
For networked sensors that report binary local decisions to a fusion center, we use a fusion rule that employs the summation of these local decisions for hypothesis testing. Based on the assumption that the received signal power decays as the distance from the target increases, exact system level detection performance measures are derived analytically. The evaluation of the probability of detection involves multiple-fold integrations. Two approximations of the probability of detection, by using binomial distribution with or without ignoring the border effect of the region of interest (ROI), are presented. It is shown that for various system parameters we have explored, the approximation that takes into account the border effect provides a very accurate estimation of the probability of detection. To achieve a better system level detection performance, the local sensor level decision threshold is chosen such that it maximizes the Kullback Leibler distance of the distributions conditioned on the two hypotheses.Keywords
This publication has 6 references indexed in Scilit:
- Saddlepoint Approximation for Sensor Network Optimization2006
- Distributed Detection and Fusion in a Large Wireless Sensor Network of Random SizeEURASIP Journal on Wireless Communications and Networking, 2005
- Elements of Information Theory2001
- Optimal data fusion of correlated local decisions in multiple sensor detection systemsIEEE Transactions on Aerospace and Electronic Systems, 1992
- Optimum multisensor fusion of correlated local decisionsIEEE Transactions on Aerospace and Electronic Systems, 1991
- Optimal Data Fusion in Multiple Sensor Detection SystemsIEEE Transactions on Aerospace and Electronic Systems, 1986