一种分布式检测系统的递推反馈自适应学习算法

Recursive feedback adaptive algorithm for decentralized detection system

分布式多传感器检测系统中的等概率假设在检测概率未知及时变情况下不能得到最优的检测状态．研究分布式检测系统的最优检测问题,考虑传感器虚警与漏报的概率未知,且概率不相等的情况,提出了一种递推的状态反馈自适应学习算法,通过在线的修正融合权值,最终使系统收敛于最佳权值,并对算法收敛性和误差进行了理论分析．仿真研究了概率未知、环境时变等情况下的算法性能,验证了所提方法的有效性．

Most decentralized multisensor detection systems, employing equal probability hypothesis, are unable to keep the optimal detection status when the detection probability is unknown or varying. The problem of optimal detection problem of the decentralized detection system is considered in this paper. Firstly, a recursive state feedback adaptive algorithm is developed when the sensor＇s false alarm probability and miss alarm probability are unknown and unequal. Based on the online correcting fusion weights, the weights will converge to the optimal values. The convergence and the steady error are then analyzed. The effects of unknown probability and variance variety on the environment of the approach are also analyzed. Finally, simulation results are given to confirm that the performance of the proposed fusion algorithm is satisfactory.