基于贝叶斯准则的硬判决融合协同频谱感知最优化  

Optimization for H-D information fused cooperative spectrum sensing based on Bayesian criterion

作　　者：井俊[1] 徐友云[1] 马文峰[1] 

机构地区：[1]解放军理工大学通信工程学院,江苏南京210007

出　　处：《电路与系统学报》2011年第1期31-39,共9页Journal of Circuits and Systems

基　　金：国家"973"资助项目(2009CB3020402);国家"863"资助项目(2009AA01Z249)

摘　　要：将硬判决融合协同频谱感知描述为贝叶斯二元假设检验问题,本文考虑感知信息传输错误的可能性,以最小化平均判决风险(贝叶斯风险)为目标的最优本地判决和最优判决融合可分别归结为LRT(likelihood ratio test)问题,并证明基于能量检测的本地LRT与观测量的门限判决等价。当仅有本地判决结果可用时,融合中心通常假设本地观测量独立同分布,可证明此时的最优融合准则为N中取K的投票准则,并给出一种低复杂度的数值迭代算法来求解最优本地判决门限和投票融合门限。数值结果显示,大多数情况下最优N中取K的投票融合可被多数逻辑融合代替而几乎不增加判决风险。H-D （Hard-Decision） information fused cooperative spectrum sensing can be formulated as a Bayesian binary hypothesis testing problem. With the sensing information transmission error probability included in the optimization for minimizing overall average decision risk （Bayes risk）, the optimal local decision rule and the optimal decision fusion rule can be expressed as LRT （likelihood ratio test） problems, respectively. When only local decisions are available to the fusion center, independent and identically distributed local observations are usually assumed, and the optimal fusion rule is proved to be equal to a simple K out of N voting rule. A low-complexity iterative algorithm then can be constructed to pursue the optimal voting threshold and local decision threshold. Numerical results show that the optimal K out of N voting fusion can be replaced by majority logic fusion in most cases without additional decision risk.

关 键 词：认知无线网络 协同频谱感知 硬判决融合 贝叶斯最优化 

分 类 号：TN929.5[电子电信—通信与信息系统]