摘要： 异常事件检测问题是无线传感器网络中的研究热点之一。为提高检测效率，提出一种基于压缩感知的异常事件检测方案。通过压缩采样得到各个节点感知数据的测量值，将异常事件检测问题建模为带权的l1范数最小化问题，采用正交匹配追踪算法进行迭代求解，根据检测函数对求解结果进行判断，并依据判断结果更新权值，开始下一轮迭代，直到检测出无线传感器网络中存在的所有异常事件。仿真实验结果表明，该方案的漏检率和误警率较低，与CCM和GEP-ADS方案相比，分别能节省约4.1%和5.8%的能耗。

关键词: 无线传感器网络, 异常事件检测, 压缩感知, 测量值, 迭代, 权值

Abstract: The anomaly event detection problem in Wireless Sensor Network(WSN) is currently a hot topic. In order to improve the detection efficiency, this paper proposes an anomaly event detection scheme based on compressive sensing. The measurements of the sensed data are obtained based on the compressive sampling, and the anomaly event detection problem is modeled as the reweighted l1 minimization problem, which is iteratively solved by the Orthogonal Matching Pursuit(OMP) algorithm. Furthermore, the solution is judged by the detection function. The weight is refreshed in the next iteration according to the judgments, until all abnormal events are detected in Wireless Sensor Network(WSN). Experimental results show that the proposed scheme can obtain the lower probability of missed detection and false alarm in different noise environments. Compared with the CCM and GEP-ADS scheme, the energy consumption of this scheme id saved by approximately 4.1% and 5.8%.

Key words: Wireless Sensor Network(WSN), anomaly event detection, compressive sensing, measurement, iteration, weight

中图分类号: 

• TP393