基于LSTM的调制干扰信号识别算法研究  

Research on Modulation Interference Signal Recognition Algorithm Based on LSTM

作　　者：王宇鹏 马晨牧 樊馨月 WANG Yu-peng;MA Chen-mu;FAN Xin-yue(Liaoning General Aviation Research Institute,Shenyang 110136,China;School of electronic information engineering,Shenyang University of Aeronautics and Astronautics,Shenyang 110136,China)

机构地区：[1]辽宁通用航空研究院,辽宁沈阳110136 [2]沈阳航空航天大学电子信息工程学院,辽宁沈阳110136

出　　处：《电脑与信息技术》2021年第4期48-52,共5页Computer and Information Technology

基　　金：辽宁省自然科学基金(项目编号:2019-ZD-0220);辽宁省教育厅重点攻关项目(项目编号:JYT19008);中国民用航空局2019年度民航安全监管能力建设项目。

摘　　要：针对常用机器学习算法对各种调制干扰信号的特征提取困难和识别率低的问题,提出了一种基于长短时记忆神经网络的通信信号调制识别方法。该算法构造了两级级联LSTM神经网络,并采用两层全连接层与LSTM层交替连接的方法,逐层提取干扰信号特征;并采用相关模型优化策略对算法所训练模型进行优化,最后相对较高地识别出各种调制干扰信号。此外,为对多方式干扰进行信号鉴别,也加入了CDMA信号的识别。结果表明,在信噪比大于5dB时,模型对各种调制信号的平均识别准确率在95%以上。In view of the difficulty of feature extraction and low recognition rate of various modulation interference signals by common machine learning algorithms,a modulation recognition method for communication signals based on long and short time memory neural network was proposed.This algorithm constructs a two-level cascading LSTM neural network,and adopts the method of alternating connection between two full connection layers and LSTM layers to extract the features of the interference signal layer by layer.The training model is optimized by using relevant model optimization strategy,and finally various modulation interference signals are identified relatively high.In addition,CDMA signal recognition is added for signal identification of multimode interference.The results show that the average recognition accuracy of the model is more than 95%when the SNR is greater than 5dB.

关 键 词：调制干扰信号 长短时记忆 信号特征 信号鉴别 优化策略 

分 类 号：TN911[电子电信—通信与信息系统]