Research On Complex Event Detection In Wireless Sensor Networks

Wireless sensor networks connect the physical world and the world ofinformation together, has changed the way people get information.With thedevelopment of microelectronic technology, radio technology and sensortechnology, the application of wireless sensor networks will become increasinglywidespread.In applications of wireless sensor networks, complex event detection isa major problem,attracting enough attention of the researchers.Because of datastorage, computing and energy, the traditional event detection algorithm can not bedirectly applied to sensor networks.Thus, it is important to design effectivealgorithms to perform event detection with low cost and low communicationoverhead without the loss of accuracy.In order to achieve effective complex event detection,this paper studies theexisting complex event detection algorithm in the field of wireless sensor networksand other areas.First of all,based on wireless sensor network characteristics and theway of process complex event data,complex events is divided into three modes:pattern matching modal, time-frequency conversion modal and data gatheredmodal.This paper studies the first two modals complex event detection.We definecomplex event model in each modal. And design complex event detection algorithmunder different modes, respectively.For the pattern matching modal, we havedesigned a complex event detection algorithm based on pattern matching andcorrelation analysis.Event was detected by the performance of single node as wellas sensor node data correlation. We reduce network traffic by transform thecorrelation from the data space into the feature space.In the time-frequencyconversion modal,we transform time domain signal into the frequency domainsignal by calculating power spectrum and then extract feature information.Tosimplify operation,we using the histogram to extract feature.While multi-modalcomplex event detection,we combine the push and back algorithm to reducenetwork traffic.The experimental results shows the proposed algorithm caneffectively guarantee the detection efficiency and reduce network traffic.