基于改进型支持向量机的频谱感知算法  

Research of spectrum sensing algorithm based on modified support vector machine

作　　者：刘晓乐[1] 张黎烁[1] 龚志恒[2] 王鑫[2,3] 

机构地区：[1]河南工程学院计算机学院,河南郑州451191 [2]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168 [3]东北大学信息科学与工程学院,辽宁沈阳110004

出　　处：《计算机工程与设计》2014年第9期3003-3006,3178,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61301232);河南省科技厅基础与前沿技术研究计划基金项目(142300410131)

摘　　要：针对无线信道环境中,信道多径衰落和噪声不确定性等低信噪比情况下主用户信号检测性能较低的问题,提出一种基于改进型支持向量机(support vector machine,SVM)的主用户信号频谱感知算法。对信号循环平稳特征参数进行特征提取,作为训练样本和待测样本;采用改进的SVM算法分别对有无主用户情况下的信号进行分类检测。仿真结果表明,与能量检测法(ED)和循环平稳特征检测法(CD)相比较,该算法可在低信噪比情况下不受噪声不确定性等因素影响,具有较高的分类检测性能,有效地实现了对主用户信号的感知。Aiming at the low accuracy rate of the primary user detection in the wireless channel environment which caused by channel multipath fading and noise uncertainty, a method was proposed based on modified support vector machine （MSVM） for the primary user spectrum sensing of cognitive radio environment in the case of low SNR. Cyclostationary characteristic parameters were extracted as training samples and testing samples, then the MSVM was trained through the samples. Finally, the trained MSVM was utilized to detect the primary user. The experimental results show that the proposed algorithm is not affected by uncertainty factors of noise and better performs in terms of classification detection compared with energy detection （ED） and cyclostationary feature detection （CD）.

关 键 词：认知网络 支持向量机 频谱感知 循环谱特征 分类检测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术] TN92[自动化与计算机技术—计算机科学与技术]