摘要
自动调制识别(Automatic Modulation Recognition,AMR)是复杂电磁环境下的信号感知和识别领域中的重要技术,广泛应用于频谱感知、链路自适应、干扰防护等领域。传统的AMR方法主要依赖于人工提取特征、决策理论和识别器的选择。而深度学习(Deep Learning,DL)算法直接从海量数据中自动获取信号特征,同时实现特征提取和识别。因此,针对复杂多变的电磁环境中的信号识别问题,提出了将DL算法应用于AMR任务。首先,从数据集、信号表示和网络模型三个层面系统地综述基于DL的AMR方法;其次,详细总结了针对不同的信号表示所设计的神经网络模型,其中接收信号可以由专家特征、序列和图像来表示;最后概述了AMR存在的问题、潜在的研究方向和结论。
Automatic modulation recognition(AMR)is an important technology for signal sensing and recognition in complex electromagnetic environment.It is widely used in civil and military fields including spectrum sensing,link adaptation,interference protection and so on.Traditional AMR methods mainly rely on manual feature extraction,decision theory and selection of recognizer.Deep learning(DL)algorithm automatically obtains signal features directly from massive data,and realizes feature extraction and recognition at the same time.Therefore,aiming at the problem of signal recognition in complex and changeable electromagnetic environment,DL algorithm is applied to AMR task.Firstly,an AMR method based on DL is systematically summarized from three aspects:data set,signal representation and network model.Secondly,neural network models designed for different signal representations are summarized in detail,in which the received signals can be represented by expert features,sequences and images.Finally,existing problems,potential research directions and conclusions of AMR are summarized.
作者
张茜茜
王禹
林云
桂冠
ZHANG Xixi;WANG Yu;LIN Yun;GUI Guan(College of Telecommunications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)
出处
《无线电通信技术》
2022年第4期697-710,共14页
Radio Communications Technology
基金
科技创新2030—“新一代人工智能”重大项目(2021ZD0113003)。
关键词
自动调制识别
信号感知
深度学习
automatic modulation recognition(AMR)
signal sensing
deep learning(DL)
