电子学报 ›› 2020, Vol. 48 ›› Issue (3): 456-462.DOI: 10.3969/j.issn.0372-2112.2020.03.006

• 学术论文 • 上一篇    下一篇

基于扩张残差网络的雷达辐射源信号识别

秦鑫, 黄洁, 查雄, 骆丽萍, 胡德秀   

  1. 中国人民解放军战略支援部队信息工程大学, 河南郑州 450001
  • 收稿日期:2019-04-02 修回日期:2019-08-26 出版日期:2020-03-25
    • 作者简介:
    • 秦鑫 女,1994年出生,重庆人.现为战略支援部队信息工程大学研究生,主要研究方向为雷达信号处理.E-mail:qinxin_0920@163.com;查雄 男,1995年出生,江西九江人.现为现为战略支援部队信息工程大学研究生,主要研究方向为智能信号处理.;黄洁 女,1973年出生,河南郑州人.现为战略支援部队信息工程大学教授、硕士生导师,主要研究方向为信息融合、模式识别.;骆丽萍 女,1978年出生,四川新津人.现为战略支援部队信息工程大学讲师,主要研究方向雷达信号处理、模式识别.
    • 基金资助:
    • 国家自然科学基金 (No.61703433)

Radar Emitter Signal Recognition Based on Dilated Residual Network

QIN Xin, HUANG Jie, ZHA Xiong, LUO Li-ping, HU De-xiu   

  1. PLA Strategic Support Force Information Engineering University, Zhengzhou, Henan 450001, China
  • Received:2019-04-02 Revised:2019-08-26 Online:2020-03-25 Published:2020-03-25
    • Supported by:
    • National Natural Science Foundation of China (No.61703433)

摘要: 针对低信噪比条件下,复杂多类雷达辐射源信号识别存在特征提取困难,识别正确率低的问题,本文提出了一种基于时频分析和扩张残差网络的辐射源信号自动识别方法.首先通过时频分析将信号时域波形转换成二维时频图像以反映信号本质特征;然后进行时频图像预处理以保留时频图像完备信息,适应深度学习模型输入;最后构建扩张残差网络以自动提取信号时频图像特征,实现雷达辐射源信号分类识别.实验结果表明,信噪比为-6dB时,该方法对16类雷达辐射源信号的整体识别正确率能够达到98.2%,对时频图像特征相似的类LFM(Linear Frequency Modulation)信号的整体识别正确率超过95%.本文提供了一种新的雷达辐射源信号智能识别方法,具有较好的工程应用前景.

关键词: 新体制雷达, 雷达信号识别, 时频分析, 图像预处理, 深度学习, 扩张残差网络

Abstract: This paper proposes a radar emitter signal recognition method based on time-frequency analysis and dilated residual network (DRN) to solve the problem of difficulty in feature extraction and low accuracy in recognition of complex multiple radar emitter signals under low signal-to-noise ratio (SNR). Firstly, the signal time-domain waveform is transformed into a two-dimensional time-frequency image by time-frequency analysis to reflect the essential characteristics of signal. Then the time-frequency image pre-processing is carried out to retain the time-frequency image complete information and adapt to the deep learning model input. Finally, the DRN is constructed to automatically extract the signal time-frequency image features and realize the recognition of radar emitter signal. Experimental results show that when the SNR is -6dB, the overall recognition rate of the proposed method for 16 types of radar signals can reach 98.2%, and the overall recognition rate for time-frequency image similar to linear frequency modulation (LFM) signals is more than 95%. In this paper, a new intelligent recognition method for radar emitter signal is presented, which has nice engineering application prospects.

Key words: new system radar, radar signal recognition, time-frequency analysis, image pre-processing, deep learning, dilated residual network

中图分类号: