Research On Distributed Detection Fusion In Wireless Sensor Networks

Distributed detection fusion is an important research content of information fusion.In the distributed detection networks,each sensor makes a decision based on its own observation,and then sends it to the fusion center.The fusion center performs a hypothesis test based on the decisions of all the sensors,and then makes a global decision.However,the limited energy,the limited computing capacity,the communication constraint,the fading channels and transmission errors in wireless sensor networks challenge the theory of distributed detection fusion.The detection fusion technologies in wireless sensor networks are studied in this dissertation and the innovations are listed as follows.1.In resource-constrained wireless sensor networks,the three-level quantization strategy is introduced,and the optimal likelihood ratio fusion rule based on instantaneous channel state information is derived.Then,three suboptimal fusion rules with the knowledge of channel statistics are proposed under Rayleigh,Rice and Nakagami fading channels,respectively.Under the condition of low signal-to-noise ratio,two reduced forms are also derived.Through the simulation,the effectiveness of the reduction on information transmission is verified.2.The distributed detection problem under grouped multiple-access channels is investigated.Firstly,for the partial coherent detection,the linear-combining fusion rules whose combining weights are chosen to maximize the deflection coefficient of the fusion statistic are proposed.Then for the noncoherent detection,the optimal likelihood ratio fusion rule and the simplified linear combination fusion rule are derived.Moreover,when the channel signal-to-noise ratio is high,the two-step fusion rule is derived.Under the condition of uniform grouping,the asymptotic behavior of the simplified fusion rules is analyzed.Through simulation,the system performance is evaluated.3.The distributed detection of weak signals in the context of multiplicative noise is investigated.Firstly,aiming at distributed detection under the ideal channel scenario,the Rao fusion rule based on one-bit quantization is derived.The system performace of distributed detection is compared with that of centralized detection.Then,for the Rayleigh fading channel,the Rao fusion rules based on instantaneous channel state information and channel statistics are derived,respectively.The quantization threshold of each sensor is optimized by using asymptotic distribution of the proposed fusion statistics.Finally,the theoretical analysis results are verified by simulation.4.The distributed detection fusion problem based on signal attenuation model is investigated.Firstly,for the coherent detection,the fusion statistic based on generalized locally most powerful(GLMP)test is derived and its asymptotic behavior is analyzed.Another GLMP fusion statistic based on the received signal power is then derived for the noncoherent detection.By optimizing the asymptotic distribution parameters of the fusion statistics,the optimal sensor decision threshold is obtained.Finally,the simulation and analysis are carried out.The performance of GLMP fusion statistic is compared with that of generalized likelihood ratio fusion statistic,and the effectiveness of the optimization method is verified.