基于监督学习的自适应协作频谱感知算法

An adaptive cooperative spectrum sensing algorithm based on supervised learning

在认知光通信网络中,能量检测方法因不需要知晓主用户的参数而被广泛研究,但该方法在噪声波动情形下的感知性能有待提高。该文提出认知用户根据光通信环境变化自动调整检测阈值的方法来提升低信噪比下的感知性能。融合中心应用坐标搜索算示为认知用户提供最优控制参数。认知用户以中心阈值为基础,依据最优参数及自身收集的能量动态设定检测阈值,并通与融合中心的信息交换,学习特定环境下的最佳阈值。此外,参考了各认知用户在协作感知中的贡献差异,并设计新的权值计算方法来体现该差异。仿真实验结果表明该文中的频谱感知方法对噪声波动有卓越的鲁棒性,提高检测概率的同时降低了错误概率,在信噪比为-20dB时的检测概率仍高达96.9%,远高于传统方法。

Primary user （PU）, In order to improve sensing performance of traditional energy detectionunder low signal-to-noise ratio （SNR）,a new method is proposed,in which cognitive user automatically adjusts the detection threshold based on changes in the radio environment. Fusion center applies coordinate search algorithm to provide optimal control parameter for the cognitive user. Based on center threshold, cognitive user dynamic sets the detection threshold according to the optimal parameter and energy collected by itself,and learns the optimal threshold for specific radio environment by exchanging information with the fusion center. In addition, differences in sensing contribution of cognitive users are fully considered and a new weighting calculation method is designed to reflect the differences. Experimental results show that the proposed spectrum sensing method has excellent robustness to noise fluctuation,and can improve detection probability and reduce error probability simultaneously. The detection probability is as high as 96.9 % while SNR is -20dB, which is much higher than those of other methods.