基于深度学习的动态主用户频谱感知算法  

Dynamic primary user spectrum sensing algorithm based on deep learning

作　　者：李新玉 赵知劲[1,2] Li Xinyu;Zhao Zhijin(School of Communication Engineering,Hangzhou Dianzi University,Hangzhou 310018,China;National Key Laboratory of Communication System Information Control Technology,36th Research Institute of China Electronics Technology Group,Jiaxing 314001,China)

机构地区：[1]杭州电子科技大学通信工程学院,浙江杭州310018 [2]中国电子科技集团第36研究所通信系统信息控制技术国家级重点实验室,浙江嘉兴314001

出　　处：《电子技术应用》2024年第1期60-65,共6页Application of Electronic Technique

基　　金：国家自然科学基金(U19B2016);浙江省教育厅一般科研项目(Y202249757)。

摘　　要：实际的频谱感知场景中主用户可能随机到达或者离开,当主用户状态在实时频谱感知期间动态变化时,现有的静态频谱感知算法性能急剧恶化。针对该现状,研究提出基于残差收缩注意力机制的动态主用户频谱感知算法。频谱感知间隔内,主用户随机到达或者随机离开的时间服从均匀分布。采用深度残差收缩网络(DRSN)提取动态主用户特征,并且滤除冗余的噪声特征;利用协调注意力模块(CAM)增强每个通道不同方向的特征信息,提高模型对动态主用户特征的表达能力。仿真结果表明,所提算法性能优于对比算法ResNet、CBAM_IQ和CBAM_Energy,所提算法对主用户随机到达或者离开服从不同分布的主用户都可以保持较高的检测概率。In actual spectrum sensing scenarios,the primary user may arrive or leave randomly,and when the primary user state changes dynamically during real-time spectrum sensing,the performance of the existing static spectrum sensing algorithm deteriorates sharply.For this situation,this paper propose a dynamic primary user spectrum sensing algorithm based on the residual shrinkage and attention mechanism.During the spectrum-sensing interval,the time when the primary user randomly arrives or leaves randomly follows a uniform distribution.The“deep residual shrinkage network(DRSN)”is used to extract dynamic primary user features and filter out redundant noise features.The“coordination attention module(CAM)”is used to improve the ability of the model to express the features of the dynamic primary user.Simulation results show that the proposed algorithm performs are better than ResNet algorithm,CBAM_IQ algorithm and CBAM_Energy algorithm.The proposed algorithm can maintain a high detection probability for the primary users who randomly arrive or leave following different distributions.

关 键 词：认知无线电 频谱感知 动态主用户 深度残差收缩网络 协调注意力机制 

分 类 号：TN925[电子电信—通信与信息系统]