基于神经网络的认知无线电合作频谱感知  

Cooperative spectrum sensing based on neural network in cognitive radio

作　　者：薛建伟 唐亮[1,2] 卜智勇 XUE Jianwei;TANG Liang;BU Zhiyong(Shanghai Institute of Microsystem and Information Technology,Chinese Academy of Sciences, Shanghai 200050, China;University of Chinese Academy of Sciences, Beijing 100049, China;School of Information Science & Technology,ShanghaiTech University, Shanghai 201210, China)

机构地区：[1]中国科学院上海微系统与信息技术研究所,上海200050 [2]中国科学院大学,北京100049 [3]上海科技大学信息学院,上海201210

出　　处：《中国科学院大学学报（中英文）》2020年第3期379-386,共8页Journal of University of Chinese Academy of Sciences

基　　金：中国科学院上海微系统与信息技术研究所“新微之星”项目(Y86QDA1001)资助。

摘　　要：认知无线电是一种可用于有效缓解当前频谱资源紧张的技术,而频谱感知是认知无线电的前提。针对低信噪比情况下频谱感知性能差的问题,提出一种将信号高阶统计量、协方差矩阵特征值与神经网络相结合的合作频谱感知算法。该算法考虑到认知用户与授权用户的信道衰落情况,利用神经网络较强的多分类能力,将最大-最小特征值之比、平均-最小特征值之比以及高阶统计量作为特征参数,通过神经网络实现合作频谱感知。仿真结果表明,该算法不仅在低信噪比情况下较其他算法具有更高的频谱检测率,而且对频谱中信号的调制类型也有较高的识别率。Cognitive radio is a technology that can be used to effectively alleviate the current strain of spectrum resources,and spectrum sensing is the prerequisite of cognitive radio.To overcome the poor performance under the condition of low SNR,a cooperative spectrum sensing algorithm combining the high-order cumulants of signal and the eigenvalue of covariance matrix with the neural network is proposed.The algorithm takes into account the channel fading between the cognitive users and primary users and utilizes the strong multi-classification ability of neural networks.The ratio of maximum-minimum eigenvalues,the ratio of average-minimum eigenvalues,and high-order cumulants are used as the inputs of neural networks to realize the spectrum sensing.The simulation results show that the proposed algorithm not only has higher spectrum detection rate than other algorithms at low SNR,but also identifies the modulation type of the signal.

关 键 词：认知无线电 频谱感知 神经网络 

分 类 号：TN929[电子电信—通信与信息系统]