The Research For Distributed Detection In Wireless Sensor Network

This paper is based on target detection in wireless sensor network.The main work is to analyze the performance of distributed detection and to introduce several distributed detection methods in wireless sensor network.The methods are introduced as follows:A neural network is trained for signal classification in a parallel topology in order to detect whether a target present or not;Serial distributed detection method in wireless sensor network when the channel noise is concerned;Serial distributed detection method in wireless sensor network when the distances between target and sensor nodes are considered in signal attenuation.Since the likelihood ratio test based optimal distributed detection in wireless sensor networks requires information about the channel state,noise distribution and sensors' detection performance,it can hardly meet the requirements of the practical applications. The support vector machine is adopted to perform distributed detection in wireless sensor networks so that the dependence of the decision method on the channel state and noise distribution information can be avoided,and the function of the wireless sensor node can also be simplified.Each sensor node retransmits an amplified version of its measurement to the data fusion center,and the data fusion center makes a final decision by means of a well trained support vector machine.The simulation results show that the support vector machine based distributed detection method in wireless sensor network has a very high detection performance.Since the channel noise in wireless sensor network is hardly avoided in practical, the channel noise is taken into account in serial distributed detection in this paper. Different from the parallel distributed detection,in serial distributed detection,local results are transmitted to the fusion center through multi-hop,short-range communications.The simulation results show that the short-range,multi-hop serial distributed detection can greatly reduce the energy consumption in detection when compared with parallel scheme,though the performance of serial distributed detection is inferior to that of parallel distributed detection.The effect of the distances between target and sensor nodes is concerned in signal detection in this paper.Since the furtherer from the target,the weaker the signal is,we divided the detection region into several smaller units.So that the serial signal detection can be made through a shortest routing in each smaller region and the result of each region is transmitted to the fusion center through a shortest path.The simulation results show that the performance of this method can reach to a certain level and the system energy consumption is reduced as well.