Np-based Guidelines For The Parallel Distributed Detection Theory

The distributed detection based on N-P criteria with parallel network is studied in this paper.First, the studying background and significance of this paper is analyzed briefly, the present research methods in distributed detection based on N-P criteria are discussed. The research content of this dissertation is introduced finally.In chapter 2, the theory of distributed detection with independent condition is studied. Based on N-P criteria, two algorithms of the weights of LDs (local detector) are decided, the first one is the algorithm of on-line weight learning based on K-L distance and the second one is the algorithm of weight self-learning based on steepest descent method, both the two algorithms nearly solved the decision problem of the weights of LDs when the knowledge of LDs is lacking. The validity of the two algorithms is validated by computer simulation.In chapter 3, the theory of distributed detection with correlated condition is studied. Firstly, the optimal fusion rule is analyzed when the joint pdf is known. Secondly, when the joint pdf is unknown, empirical estimation is adopted to solve this problem, the relationship of the training sample size and the estimated confidence is analyzed.The problems and content needed further research are pointed out in chapter 4.