中国空间科学技术

中国空间科学技术 ›› 2023, Vol. 43 ›› Issue (4): 126-136.doi: 10.16708/j.cnki.1000-758X.2023.0061

• 论文 • 上一篇    下一篇

基于多域融合的空间辐射源细微特征识别

王晓晗1，2，闫毅1，范亚楠1，李雪1，牟娇1   

  1. 1 中国科学院 国家空间科学中心，北京100190
    2 中国科学院大学，北京100049
  • 出版日期:2023-08-25 发布日期:2023-07-18

Space emitter fine feature identification based on multi-domain fusion

WANG Xiaohan1，2，YAN Yi1，FAN Yanan1，LI Xue1，MOU Jiao1   

  1. 1 National Space Science Center，CAS，Beijing 100190，China
    2 University of Chinese Academy of Sciences，Beijing 100049，China
  • Published:2023-08-25 Online:2023-07-18

PDF (PC)

144

摘要： 辐射源识别主要关注辐射源的个体差异，通过信号处理手段，获取辐射源信号上与设备硬件相关的特征参数，从而掌握辐射源设备的型号、工作状态等信息。由于通信信号愈发复杂，单一域特征识别无法全面反映信号的细微差异，直接对信号处理结果进行识别存在大量数据冗余，导致识别效果不佳。为提高空间辐射源的识别效果，提出了一种基于多域特征融合的辐射源识别方法，通过对信号进行HilbertHuang变换和高阶谱分析，并提取变换结果的均值、方差、峰度、偏度和信息熵，将其拼接、融合形成空间辐射源信号的多域特征，利用SVM分类器进行分类，实现多域融合的辐射源细微特征识别。仿真结果表明，使用多域融合方法进行辐射源识别，在20dB的信噪比下可达到95.44%的正确率，与传统基于边际谱信息熵的分类识别方法对比，正确率提升9.41%。对比基于HHT边际谱、边际谱信息熵、双谱投影、双谱矩形积分、四阶累积量切片谱的分类识别方法，本方法的识别效果均有提升。

关键词: 空间通信, 特征融合, 空间辐射源识别, Hilbert-Huang变换, 高阶谱分析

Abstract: Emitter identification mainly focuses on individual differences of emitters，and obtains characteristic parameters related to equipment hardware on emitter signals through signal processing means，so as to master the model and working status of emitter equipment.As communication signals become more and more complex，feature identification in a single domain cannot fully reflect the fine differences of signals，and there is a large amount of data redundancy in direct identification of signal processing results，resulting in poor identification effect.In order to improve the identification effect of space emitter，an emitter identification method was proposed based on the multi-domain feature fusion.By perfroming Hilbert-Huang transform and higher order spectrum analysis of the signal，the mean，variance，kurtosis，skewness and information entropy of the teansformed results are exteacted，SVM classifier is used to classify emitters and realize the fine feature identification of multidomain fusion.The simulation results show that the accuracy of multidomain feather fusion method can reach 95.44% under the SNR of 20dB，and that the accuracy is improved by 9.41% compared with the traditional classification method based on marginal spectrum information entropy.Compared with the classification and identification methods based on HHT marginal spectrum，marginal spectrum information entropy，bispectral projection，bispectral rectangular integral and fourth-order cumulant slice spectrum，the identification effect of the proposed method is improved.

Key words: space communication, feature fusion, space emitter identification, Hilbert-Huang transform, higher order spectra