电子学报 ›› 2022, Vol. 50 ›› Issue (6): 1336-1343.DOI: 10.12263/DZXB.20210882

所属专题: 电磁频谱智能+

• 电磁频谱智能+ • 上一篇    下一篇

基于多重分形谱智能分析的卫星信号调制识别研究

杨伟超, 杜宇, 文伟, 侯舒维, 徐常志, 张建华   

  1. 中国空间技术研究院西安分院,陕西 西安 710100
  • 收稿日期:2021-07-12 修回日期:2021-12-29 出版日期:2022-06-25
    • 作者简介:
    • 杨伟超 男,1983年出生,河南林州人.现为中国空间技术研究院西安分院高级工程师.主要研究方向为航天电子侦察、智能电磁频谱战等.E-mail: yangwc@cast504.com
      杜 宇 女,1987年出生,黑龙江哈尔滨人.现为中国空间技术研究院西安分院工程师.主要研究方向为高速数传、通信信号处理等.E-mail: duy@cast504.com
    • 基金资助:
    • 军委科技委基础加强计划 (军173)重点基础研究项目 (2020-JCJQ-ZD-071)

Modulation Recognition of Satellite Communication Signal Based on Intelligent Analysis of Multi-Fractal Spectrum

YANG Wei-chao, DU Yu, WEN Wei, HOU Shu-wei, XU Chang-zhi, ZHANG Jian-hua   

  1. China Academy of Space Technology(Xi’an), Xi’an, Shaanxi 710100, China
  • Received:2021-07-12 Revised:2021-12-29 Online:2022-06-25 Published:2022-06-25
    • Supported by:
    • Key Basic Research Program of CMC Science and Technology Commission Foundation Enhancement Project (Military 173 Project) (2020-JCJQ-ZD-071)

摘要:

调制方式识别是电磁频谱战中的关键技术之一,已有星上识别方法智能化程度低、适应性差.针对此类问题,提出了一种基于多重分形谱和深度学习相结合的智能识别方法.首先分析了常见卫星通信信号的多重分形特性,构建了多重分形特征域矩阵.在此基础上,将该特征矩阵与深度学习残差网络相结合,并根据多尺度思想对残差网络结构进行了优化改进,改进后残差网络的多层自主细节特征提取优势完美契合了多重分形谱多尺度特征刻画能力,最终实现了卫星通信信号调制方式的有效识别.仿真结果表明,该方法具有较好的识别性能,当信噪比不低于1 dB时,平均识别率大于89%.

关键词: 卫星通信, 深度学习, 调制识别, 多重分形谱

Abstract:

Modulation recognition is one of the key technologies in satellite communication anti-interference and interference analysis. The existing on-board recognition methods have low intelligence degree and poor adaptability. In order to solve these problems, an intelligent recognition method based on multi-fractal spectrum and deep learning is proposed. Firstly, the multi-fractal spectrum characteristics of common satellite communication signals are analyzed, a multi-fractal eigendomain matrix is constructed. On this basis, the eigendomain matrix is combined with the deep learning residual network, and the structure of the residual network is optimized and improved according to the multi-scale idea, and the multi-layer autonomous detail feature extraction advantage of the improved residual network perfectly corresponds to the multi-scale feature characterization capability of the multi-fractal spectrum. Finally, the modulation of satellite communication signal is effectively recognized. The simulation results show that this method has good recognition performance, when the SNR is not lower than 1 dB, the average recognition rate is greater than 89%.

Key words: satellite communication, deep learning, modulation recognition, multi-fractal spectrum

中图分类号: