基于1D-CNN-LSTM的特定辐射源识别方法  

Specific Emitter Identification Method Based on 1D-CNN-LSTM

作　　者：许全 谭守标[1] 孙翔[1] 樊进[2] Xu Quan;Tan Shoubiao;Sun Xiang;Fan Jin(School of Integrated Circuits,Anhui University,Hefei 230000;Network Information Center of Anhui University,Hefei 230000)

机构地区：[1]安徽大学集成电路学院,合肥230000 [2]安徽大学网络信息中心,合肥230000

出　　处：《现代计算机》2022年第12期30-34,55,共6页Modern Computer

摘　　要：特定辐射源识别(Specific Emitter Identification,SEI)是指利用雷达指纹特征确定产生信号的辐射源个体。通过对雷达辐射源的识别,可以有效区分出敌我雷达,保证雷达信息的安全性,这在电子战中具有重要的军事意义。目前传统分类识别方法存在指纹特征提取困难,指纹识别正确率低等问题。本文提出了一种基于1D-CNN-LSTM(One Dimensional Convolutional Neural Network Long Short Term Memory)特定辐射源识别方法。该方法直接使用采集到的信号的同向相交分量(Inphase/Quadrature.I/Q)数据进行信号的特征提取,并实现了对于来自不同辐射源个体信号的识别与区分。该模型兼具卷积神经网络与长短时记忆网络的优点,它可以在提取抽象特征的同时进行时序分析。实验结果表明,1D-CNN-LSTM网络能够在复杂的电磁环境下实现对特定辐射源个体的更好识别。Specific Emitter Identification(SEI)refers to the use of radar fingerprint characteristics to determine the indi⁃vidual Emitter of the signal.Through the Identification of the radar Emitter,it can effectively distinguish the enemy radar and en⁃sure the security of radar information,which has important military significance in electronic warfare At present,traditional classifi⁃cation and recognition methods are difficult to extract fingerprint features and have low correct rate of fingerprint recognition.In this paper,a method based on One Dimensional Convolutional Neural Network Long Short Term(1D-CNN-LSTM)is proposed The method directly uses the Inphase/Quadrature(I/Q)data of the collected signals to extract signal features,and realizes the rec⁃ognition and differentiation of individual signals from different radiation sources The model has the advantages of both convolu⁃tional neural network and short and long time memory network.It can extract abstract features and perform time series analysis at the same time.Experimental results show that the 1D-CNN-LSTM network can realize better identification of specific radiation sources in complex electromagnetic environment.

关 键 词：脉内无意调制 特定辐射源识别 卷积神经网络 长短时记忆网络 

分 类 号：TN957.51[电子电信—信号与信息处理]