基于渐进性能分析的最优分布式量化检测算法

One-bit quantization design for distributed detection based on asymptotic analysis

针对无线传感器网络中对发送功率和传输带宽限制的问题,提出基于规则量化(RQ)和不规则量化(NRQ)的一比特分布式检测算法。该检测方案在传感器节点处引入量化器,整个系统采用并行结构,由N个带有量化器的传感器节点组成。融合中心处采用广义似然比检验(GLRT)作为融合准则,并根据接收到的二进制信息作出全局判决。通过仿真实验对所提出的检测方案性能进行验证,并与未量化的检测方案进行比较。仿真结果表明:最优量化阈值提升了融合中心的检测概率,在理想信道和差错信道(BSC)下,基于不规则量化方案的检测系统性能明显优于基于规则量化方案的检测系统。

In view of transmission power/bandwidth constrains in wireless sensor networks（WSNs）, two distributed detection schemes, namely, 1-bit regular quantizer（RQ） and 1-bit non-regular quantizer（NRQ）, were proposed. The quantizers were employed in local sensor nodes. The decentralized detection system, adopting parallel structure, consists of N sensor nodes with quantizers. The generalize likehood ratio test（GLRT） was employed as a fusion rule at the fusion center（FC）, where a global decision was performed. The performances of the proposed schemes were verified by a simulation, and the proposed schemes were compared with each other as well as with those without quantization. The results show that the detection performance is improved by optimizing the local quantization threshold. The performance of the 1-bit GLRT detector based on NRQ scheme outperforms the counterpart based on RQ scheme, either in a perfect channel or in binary symmetric channel（BSC）.