基于改进残差网络的雷达辐射源信号识别  

Radar Emitter Signal Recognition Based on Improved Residual Network

作　　者：郭恩泽 张洪德 杨雷 刘益岑 彭镜轩 张磊 GUO Enze;ZHANG Hongde;YANG Lei;LIU Yicen;PENG Jingxuan;ZHANG Lei(Communication Sergeant School,Army Engineering University of PLA,Chongqing 400035,China;College of Communication Engineering,Army Engineering University of PLA,Nanjing 210007,China;State Key Laboratory of Blind Signal Processing,Chengdu 610041,China)

机构地区：[1]中国人民解放军陆军工程大学通信士官学校,重庆400035 [2]中国人民解放军陆军工程大学通信工程学院,江苏南京210007 [3]信号盲处理国家级重点实验室,四川成都610041

出　　处：《无线电工程》2022年第12期2178-2185,共8页Radio Engineering

摘　　要：针对现有雷达辐射源信号识别方法存在的识别率不足、网络模型复杂和鲁棒性差等问题,提出一种基于改进残差网络的雷达辐射源信号识别方法。通过时频分析方法将一维时域信号变换成二维时频图像,用作深度学习网络模型的训练和测试;借鉴轻量化和去“网格效应”的设计思想,构建改进的残差网络模型。实验结果表明,在-15~-10 dB低信噪比条件下,该模型对7种雷达信号的综合识别率为95.9%,比GoogLeNet,AlexNet,MobileNetV2模型分别高5.6%,3.1%,1.4%,与ResNet18模型相比识别率接近,复杂度大大减少。设计了一种新的用于雷达辐射源信号识别的模型,具有较好的工程应用前景。To solve the problems of insufficient recognition rate,complex network model and poor robustness of the existing radar emitter signal recognition methods,a radar emitter signal recognition method based on improved residual network is proposed.Firstly,one-dimensional time-domain signal is transformed into two-dimensional time-frequency image by time-frequency analysis method,which is used for training and testing of deep learning network model;secondly,an improved residual network model is constructed based on the design idea of lightweight and elimination of“grid effect”.The experimental results show that under the condition of-15~-10 dB low signal to noise ratio,the comprehensive recognition rate of the model for seven radar signals is 95.9%,which is 5.6%,3.1%and 1.4%higher than GoogLeNet,AlexNet and MobileNetV2 models respectively.Compared with ResNet18 model,the recognition rate is close,and the complexity is greatly reduced.A new model for radar emitter signal recognition is designed and has a good engineering application prospect.

关 键 词：雷达辐射源 深度学习 残差网络 分类识别 

分 类 号：TN911.7[电子电信—通信与信息系统]