欢迎访问《兵工学报》官方网站,今天是

兵工学报 ›› 2020, Vol. 41 ›› Issue (7): 1464-1472.doi: 10.3969/j.issn.1000-1093.2020.07.026

• 研究简报 • 上一篇    

基于联合投影字典学习的辐射源调制识别

李东瑾, 杨瑞娟, 董睿杰   

  1. (空军预警学院, 湖北 武汉 430019)
  • 收稿日期:2019-07-22 修回日期:2019-07-22 上线日期:2020-09-23
  • 作者简介:李东瑾(1992—), 男, 博士研究生。E-mail: li_dong_jin@163.com
  • 基金资助:
    国防科技创新特区基金项目(17H86304ZT00302201)

Emitter Signal Modulation Recognition Based on Joint Projection Dictionary Learning

LI Dongjin, YANG Ruijuan, DONG Ruijie   

  1. (Air Force Early Warning Academy, Wuhan 430019, Hubei, China)
  • Received:2019-07-22 Revised:2019-07-22 Online:2020-09-23

摘要: 针对字典学习用于辐射源识别时原子表征能力有限和复杂环境适应性不足问题,提出一种基于联合投影字典学习的辐射源识别方法。利用时频变换提取辐射源信号初始特征,并通过降维、降噪实现特征预处理;采用核空间投影和降维投影学习方式优化字典原子结构,基于数据集训练获取联合投影字典;通过分类测试完成了有效性验证。仿真结果表明:该方法所提取字典原子具备较强表征能力,能够适应参数多变的复杂环境;较常规有监督字典学习方式更易区分多类型、高相似度信号,-6 dB时单载频信号、线性调频信号、非线性调频信号、二相编码信号、四相编码信号、Frank信号、二相频率编码信号、四相频率编码信号、非线性调频-二相编码复合调制信号、二相频率编码-二相编码复合调制信号10类辐射源信号的整体平均识别率为94.4%.

关键词: 辐射源识别, 联合投影字典学习, 时频特征, 降维学习, 稀疏编码

Abstract: An emitter signal recognition method based on joint projection dictionary learning (JPDL) is proposed for the limited atomic representation ability and the insufficient adaptability of complex environment for dictionary learning. The initial features of emitter signal are extracted by time-frequency transform, and the feature preprocessing is realized by dimensionality reduction and noise reduction. Then the atomic structure of the dictionary is optimized by using the methods of kernel space projection and dimensionality reduction projection, and a joint projection dictionary is obtained through data set training. The validity verification is completed by the classification test. The simulated results show that the extracted dictionary atoms have strong representation ability and can adapt to the complex environment with variable parameters. Compared with the conventional supervised dictionary learning method, the proposed method can better distinguish the multi-type and high-similarity signals. The overall recognition rate of 10 types of emitter signals, such as single carrier frequency modulation (SCFM) signal, linear frequency modulation (LFM) signal, nonlinear frequency modulation (NLFM) signal, binary phase shift keying (BPSK) signal, quadrature phase shift keying (QPSK) signal, Frank signal, binary frequency shift keying (BFSK) signal, quadrature frequency shift keying (QFSK) signal, LFM-BPSK modulation signal and BFSK-BPSK modulation signal, at -6 dB is 94.4%. Key

Key words: emittersignalrecognition, jointprojectiondictionarylearning, time-frequencyfeature, dimensionalityreduction, sparsecoding

中图分类号: