摘要
面对6G频谱空间高效利用需求下动态频谱接入用户对电磁环境辐射源用频规律认知的需求,研究基于深度学习的辐射源用频行为智能识别方法。首先,提出了一种基于深度学习的浅层阈值去噪方法,克服噪声对浅层特征的影响以及非线性传播导致的对深层特征的进一步影响;然后,提出了一种基于卷积块注意力机制的用频行为识别模型,以增强对用频行为特征关键信息的关注,进一步提升识别性能。仿真结果表明,所提方法有效提升了辐射源用频规律识别的准确率和鲁棒性。
In response to the demand for efficient utilization of 6G spectrum space,and the need for dynamic spectrum access users to understand the frequency usage patterns of environmental electromagnetic radiation sources,this paper studies an intelligent identification method for frequency usage behaviors based on deep learning.Firstly,a threshold denoising method based on deep learning is proposed,which overcomes the impact of noise on shallow features and the further impact on deep features caused by nonlinear propagation.Then,a frequency usage behavior identification model based on a convolutional block attention mechanism is proposed to enhance the focus on key information of frequency usage behavior features,further improving identification performance.Simulation results show that this method effectively improves the accuracy and robustness of interference frequency usage pattern identification。
作者
姚昌华
张军
曹洪森
李洋
祁博宇
YAO Changhua;ZHANG Jun;CAO Hongsen;LI Yang;QI Boyu(School of Electronic&Information Engineering,Nanjing University of Information Science&Technology,Nanjing 210044,China;Nanjing University of Information Engineering Tianchang Research Institute,Tianchang 239300,China;College of Electronic and Information Engineering,Anhui University of Architecture,Hefei 230009,China)
出处
《移动通信》
2024年第8期46-55,共10页
Mobile Communications
基金
国家自然科学基金“复杂电磁环境下通信抗干扰智能决策技术”,“面向任务的无人机集群通控一体群智博弈学习优化研究”(U22B2002,61971439)
通信抗干扰全国重点实验室基础科研创新基金(稳定支持)项目资助“智能自组网博弈学习协同抗干扰技术研究”(IFN20230207)。
关键词
6G
频谱认知
干扰识别
深度学习
6G
spectrum cognition
interference identification
deep learning
