核支持向量的主用户活动场景分类算法  

Classification algorithm for activity scene of primary user based on kernel support vector

作　　者：张红 申滨[1] 方广进 崔太平[1] ZHANG Hong;SHEN Bin;FANG Guangjin;CUI Taiping(School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,P.R.China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2023年第1期101-109,共9页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家自然科学基金(62071078)。

摘　　要：针对认知无线电网络中传统频谱感知方法性能不足以及空白频谱利用率较低的问题,提出了一种基于核支持向量的主用户活动场景分类算法,通过判断地理区域内的活动主用户数量及分布情况来提高获得潜在频谱接入机会的可能性。根据核支持向量的边界对主用户活动场景作初分类处理,由此判定当前网络中的活跃主用户发射机的数量。初分类处理既能减少支持向量中矩阵计算量,也能减少人工标记数据所带来的成本。再对每一个初分类处理后的数据进行无监督聚类,从而得到实际对应的主用户活动场景细分类。实验结果表明,所提算法与直接使用核支持向量分类算法相比,不仅改善了频谱感知的性能,同时还大大降低了定标成本及时间成本。Aiming to solve the problems of insufficient performance of traditional spectrum sensing methods and low utilization of spectrum in cognitive radio networks,this paper proposes primary user(PUs)activity scene classification algorithm based on kernel support vector to improve the performance of spectrum sensing by judging the number of active PU transmitters(PTs)in different locations,thereby obtaining potential spectrum access opportunities.Firstly,according to the boundary of the kernel support vector,the PU activity scenes are initially classified to determine the number of PTs in the current network.The initial classification process can not only reduce the complexity of matrix calculation in the support vector,but also reduce the cost of manual labeling of data.Then,unsupervised clustering is performed after the initial classification process,so as to obtain the final actual PU activity scenes.The experimental results show that the proposed algorithm not only improves the performance of spectrum sensing,but also greatly reduces the time cost compared with the direct use of the kernel support vector classification algorithm.

关 键 词：频谱感知 机器学习 核支持向量 场景分类 

分 类 号：TN929.5[电子电信—通信与信息系统]