Compressed sensing for abnormal event detection in wireless networks

Compressed sensing (CS) is a recently developed theory that has earned increasing interests in the area of wireless communications and signal processing. It states that the main information of a signal can be recovered from a relatively small number of linear projections. Event detection is an important application of wireless networks which has also attracted much attention. Recent research shows that CS based mechanism can be applied to event detection if the signature of event is sparse in a certain domain. In this article, we propose a novel scheme for abnormal event detection in the noise-involved networking environment by an ameliorated reconstruction method, with no prior information regarding the targeted wireless networks. We also analyze the trade-off between detection probability and false alarm. Finally we give a novel metric which helps to evaluate the results of detection. Simulation shows that our scheme proves to be effective and we can also acquire the best state for appropriate detection with the global data processed.