Multilevel quantisation and fusion scheme for the decentralised detection of an unknown signal

The paper considers the problem of detecting a deterministic but unknown signal using distributed sensors. The unknown signal detection problem is considered as a hypothesis testing problem in which the alternative is quite broad. A multilevel quantisation and fusion scheme is proposed and analysed. The proposed scheme is a modified version of the counting rules and can either be thought of as a feedback of the fusion centre decision to the local sensors or as a multiple partitioning of the observation space. Based on numerical and simulation analysis, the proposed quantisation and fusion scheme is seen to be more robust than the counting rules. It performs reasonably well in most subsets of the alternative hypothesis. The performance of the scheme is also compared with that of Fisher´s combining scheme.<>