摘要
自动调制识别(Automatic Modulation Recognition,AMR)用于检测接收信号的调制样式,是通信系统进行后续信息处理的关键前提,在电子对抗、频谱管控和认知无线电等多方面获得了广泛应用。近年来深度学习(Deep Learning,DL)发展迅猛,神经元非线性变换处理和各种神经网络的灵活拼接方法使得网络模型具备较强的特征提取能力,为DL-AMR方法研究奠定了坚实基础。相较于传统的AMR方法,DL-AMR方法在识别精度和计算复杂度等方面更具优势。基于此,从AMR概览、网络模型作用机理、AMR信号模型、DL-AMR方法、开源基准数据集、模型评价指标及基线模型仿真实验六方面着手,对DL-AMR方法进行了系统综述,对研究现状进行分析总结,展望未来研究方向,进一步推动DL-AMR研究进展。
Automatic Modulation Recognition(AMR)is used to detect the modulation pattern of received signals and is a key prerequisite for subsequent information processing in communication systems.It is widely applied in various fields such as electronic warfare,spectrum control,and cognitive radio.In recent years,Deep Learning(DL)has experienced rapid development.The non-linear transformation of neurons and flexible concatenation methods of various neural networks make the network models have strong feature extraction capabilities,laying a solid foundation for the research of DL-AMR methods.Compared with the traditional AMR methods,the DL-AMR method has advantages in recognition accuracy and computational complexity.Based on this,the DL-AMR method is reviewed systematically from six aspects:overview of AMR,network model mechanism,AMR signal model,DL-AMR method,open source benchmark dataset,model evaluation indicators,and baseline model simulation experiment.The current research status is analyzed and summarized,the future research direction is prospected,which further promotes the progress of DL-AMR research.
作者
陈昊
郭文普
巨西诺
康凯
施昊
高绍原
CHEN Hao;GUO Wenpu;JU Xinuo;KANG Kai;SHI Hao;GAO Shaoyuan(College of Combat Support,Rocket Force University of Engineering,Xi'an 710025,China;Unit 96852,PLA,Shenyang 110033,China)
出处
《无线电工程》
2025年第3期526-539,共14页
Radio Engineering
