无线传感器网络中基于多比特量化的极大似然分布式估计方法  

Multi-level Quantization Scheme for Distributed Maximum Likelihood Estimation in Wireless Sensor Networks

作　　者：郭黎利[1] 高飞[1] 孙志国[1] 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《电子学报》2016年第11期2773-2779,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61271263;No.61101141)

摘　　要：在无线传感器网络背景下的分布式估计中,由于传输网络对发送功率和传输带宽的限制,压缩信源冗余、降低通信数据量便成为一个重要的课题.为此,本文提出了一种基于多比特量化观测的分布式估计方法(MQS),利用渐进性能作为优化准则构造量化阈值优化问题,运用粒子群算法对其进行求解得到最优量化阈值,给出了克拉美罗下界的解析表达式,并与均匀量化方法(UQS)和未量化方法(NQS)进行对比.理论分析和仿真实验表明,MQS的性能优于UQS.当量化深度增大到3时,MQS的估计性能十分接近NQS的估计性能.In the context of distributed estimation in wireless sensor networks （ WSN）, due to transmission power/ bandwidth constrains, it is significant to reduce size of transmitted data. In this paper, a distributed estimation scheme, name- ly, multi-level quantization scheme （MQS） is proposed. The quantization threshold optimization problem is formulated by u- sing asymptotic performance as an optimality criterion. The optimum quantization thresholds are obtained by resorting to par- ticle swarm optimization algorithm. The explicit expression of the Cram6r-Rao lower bound is derived. The proposed method is compared with uniform ＇quantization Scheme （UQS） and no quantization scheme （NQS）. Theoretical analysis and simula- tion results demonstrate that the MQS scheme outperforms the UQS. Moreover, with 3-bit quantization,the MQS can provide estimation performance very close to that of the NQS.

关 键 词：无线传感器网络 多比特量化 分布式估计 粒子群算法 

分 类 号：TN911.23[电子电信—通信与信息系统]