高级搜索

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于知识蒸馏与注意力图的雷达信号识别方法

曲志昱 李根 邓志安

曲志昱, 李根, 邓志安. 基于知识蒸馏与注意力图的雷达信号识别方法[J]. 电子与信息学报, 2022, 44(9): 3170-3177. doi: 10.11999/JEIT210695
引用本文: 曲志昱, 李根, 邓志安. 基于知识蒸馏与注意力图的雷达信号识别方法[J]. 电子与信息学报, 2022, 44(9): 3170-3177. doi: 10.11999/JEIT210695
QU Zhiyu, LI Gen, DENG Zhian. Radar Signal Recognition Method Based on Knowledge Distillation and Attention Map[J]. Journal of Electronics & Information Technology, 2022, 44(9): 3170-3177. doi: 10.11999/JEIT210695
Citation: QU Zhiyu, LI Gen, DENG Zhian. Radar Signal Recognition Method Based on Knowledge Distillation and Attention Map[J]. Journal of Electronics & Information Technology, 2022, 44(9): 3170-3177. doi: 10.11999/JEIT210695

基于知识蒸馏与注意力图的雷达信号识别方法

doi: 10.11999/JEIT210695
基金项目: 国家自然科学基金(61801143, 61971155)
详细信息
    作者简介:

    曲志昱:女,副教授,研究方向为电子侦察与对抗、阵列信号测向

    李根:男,硕士生,研究方向为雷达信号识别

    邓志安:男,副教授,研究方向为人工智能与宽带信号处理

    通讯作者:

    邓志安 dengzhian@hrbeu.edu.cn

  • 中图分类号: TN957.51

Radar Signal Recognition Method Based on Knowledge Distillation and Attention Map

Funds: The National Natural Science Foundation of China (61801143, 61971155)
  • 摘要: 针对传统雷达信号识别方法无法有效进行识别类型扩展问题,该文提出一种基于知识蒸馏与注意力图的雷达信号识别方法。首先将雷达信号的平滑伪Wigner-Ville分布(SPWVD)作为输入;然后设计了基于残差网络的增量学习网络结构,利用基于知识蒸馏与注意力图的损失函数,缓解类别增量过程中的灾难性遗忘;最后采用基于样本特征均值距离的方法对数据集进行管理,有效降低存储资源占用空间。实验表明,该方法能在存储资源有限的情况下,对扩展分类的信号快速完成训练,且对原有分类和扩展分类信号均有良好的识别准确率。
  • 图  1  雷达脉内调制信号时频图像Grad-Cam可视化效果

    图  2  基于知识蒸馏与注意力图的雷达信号识别方法流程

    图  3  增量识别卷积神经网络参数示意图

    图  4  网络结构

    图  5  基于知识蒸馏与注意力图的增量训练过程

    图  6  模型关注区域差异示意图

    图  7  准确率对比

    图  8  不同模型损失

    图  9  信噪比–5 dB识别效果对比

    图  10  联合训练与本文方法存储数量对比

    图  11  识别正确率与信噪比关系曲线

    图  12  类别扩展数量与识别正确率关系曲线

    表  1  已识别信号参数设置

    信号类型参数变化范围
    LFM初始频率$ {f_c} $
    带宽$\Delta f$
    0.01~0.45
    0.02~0.40
    Frank载频${f_0}$
    相位数
    0.10~0.45
    [4, 5, 6, 7, 8]
    BPSKBarker码
    载频${f_0}$
    码长${T_b}$
    [5, 7, 11, 13]
    0.10~0.45
    (1/32~1/16)N
    DLFM初始频率${f_c}$
    带宽$\Delta f$
    0.01~0.41
    0.05~0.45
    EQFM最小频率${f_{{\rm{min}}} }$
    最大频率${f_{{\rm{max}}} }$
    带宽$\Delta f$
    0.01~0.45
    0.01~0.45
    0.05~0.40
    SFM最小频率${f_{{\rm{min}}} }$
    最大频率${f_{{\rm{max}}} }$
    带宽$\Delta f$
    0.01~0.18
    0.20~0.45
    0.02~0.44
    下载: 导出CSV

    表  2  扩展类型信号参数设置

    信号类型参数变化范围
    2FSK载频${f_1}$,$ {f_2} $
    码宽${T_b}$
    0.10~0.45
    (1/32~1/8)N
    4FSK载频${f_1}$~$ {f_4} $
    码宽${T_b}$
    0.10~0.45
    (1/32~1/8)N
    LFM-SFM初始频率${f_c}$
    最小频率${f_{{\rm{min}}} }$
    最大频率${f_{{\rm{max}}} }$
    0.01~0.45
    0.08~0.18
    0.20~0.30
    MLFM载频${f_0}$
    带宽$ \Delta f $
    0.01~0.25
    0.10~0.45
    P1, P2载频${f_0}$
    步进频率$M$
    0.10~0.45
    [4,5,6,7,8], P2取偶数
    P3, P4载频${f_0}$
    压缩比
    0.10~0.45
    16~64
    下载: 导出CSV
  • [1] 王星, 周一鹏, 周东青, 等. 基于深度置信网络和双谱对角切片的低截获概率雷达信号识别[J]. 电子与信息学报, 2016, 38(11): 2972–2976. doi: 10.11999/JEIT160031

    WANG Xing, ZHOU Yipeng, ZHOU Dongqing, et al. Research on low probability of intercept radar signal recognition using deep belief network and bispectra diagonal slice[J]. Journal of Electronics &Information Technology, 2016, 38(11): 2972–2976. doi: 10.11999/JEIT160031
    [2] ZHANG Chunhong, HAN Yuetao, ZHANG Ping, et al. Research on modern radar emitter modelling technique under complex electromagnetic environment[J]. The Journal of Engineering, 2019(20): 7134–7138. doi: 10.1049/joe.2019.0579
    [3] 徐玉芬. 现代雷达信号处理的数字脉冲压缩方法[J]. 现代雷达, 2007, 29(7): 61–64. doi: 10.3969/j.issn.1004-7859.2007.07.018

    XU Yufen. Methods of digital pulse compression in modern radar signal processing[J]. Modern Radar, 2007, 29(7): 61–64. doi: 10.3969/j.issn.1004-7859.2007.07.018
    [4] 曲志昱, 毛校洁, 侯长波. 基于奇异值熵和分形维数的雷达信号识别[J]. 系统工程与电子技术, 2018, 40(2): 303–307. doi: 10.3969/j.issn.1001-506X.2018.02.10

    QU Zhiyu, MAO Xiaojie, and HOU Changbo. Radar signal recognition based on singular value entropy and fractal dimension[J]. Systems Engineering and Electronics, 2018, 40(2): 303–307. doi: 10.3969/j.issn.1001-506X.2018.02.10
    [5] 肖易寒, 王亮, 郭玉霞. 基于去噪卷积神经网络的雷达信号调制类型识别[J]. 电子与信息学报, 2021, 43(8): 2300–2307. doi: 10.11999/JEIT200506

    XIAO Yihan, WANG Liang, and GUO Yuxia. Radar signal modulation type recognition based on denoising convolutional neural network[J]. Journal of Electronics &Information Technology, 2021, 43(8): 2300–2307. doi: 10.11999/JEIT200506
    [6] QU Zhiyu, WANG Wenyang, HOU Changbo, et al. Radar signal intra-pulse modulation recognition based on convolutional denoising autoencoder and deep convolutional neural network[J]. IEEE Access, 2019, 7: 112339–112347. doi: 10.1109/ACCESS.2019.2935247
    [7] 郭立民, 寇韵涵, 陈涛, 等. 基于栈式稀疏自编码器的低信噪比下低截获概率雷达信号调制类型识别[J]. 电子与信息学报, 2018, 40(4): 875–881. doi: 10.11999/JEIT170588

    GUO Limin, KOU Yunhan, CHEN Tao, et al. Low probability of intercept radar signal recognition based on stacked sparse auto-encoder[J]. Journal of Electronics &Information Technology, 2018, 40(4): 875–881. doi: 10.11999/JEIT170588
    [8] 石礼盟, 杨承志, 吴宏超. 基于深层残差网络和三元组损失的雷达信号识别方法[J]. 系统工程与电子技术, 2020, 42(11): 2506–2512. doi: 10.3969/j.issn.1001-506X.2020.11.12

    SHI Limeng, YANG Chengzhi, and WU Hongchao. Radar signal recognition method based on deep residual network and triplet loss[J]. Systems Engineering and Electronics, 2020, 42(11): 2506–2512. doi: 10.3969/j.issn.1001-506X.2020.11.12
    [9] PARISI G I, KEMKER R, PART J L, et al. Continual lifelong learning with neural networks: A review[J]. Neural Networks, 2019, 113: 54–71. doi: 10.1016/j.neunet.2019.01.012
    [10] DHAR P, SINGH R V, PENG K C, et al. Learning without memorizing[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, USA, 2019: 5133–5141.
    [11] HINTON G, VINYALS O, and DEAN J. Distilling the knowledge in a neural network[J]. Computer Science, 2015, 14(7): 38–39.
    [12] SELVARAJU R R, COGSWELL M, DAS A, et al. Grad-CAM: Visual explanations from deep networks via gradient-based localization[C]. 2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 2017: 618–626.
    [13] ZHOU Bolei, KHOSLA A, LAPEDRIZA A, et al. Learning deep features for discriminative localization[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, USA, 2016: 2921–2929.
    [14] LI Zhizhong and HOIEM D. Learning without forgetting[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(12): 2935–2947. doi: 10.1109/TPAMI.2017.2773081
    [15] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, USA, 2016: 770–778.
    [16] CASTRO F M, MARÍN-JIMÉNEZ M J, GUIL N, et al. End-to-end incremental learning[C]. The 15th European Conference on Computer Vision, Munich, Germany, 2018: 241–257.
    [17] REBUFFI S A, KOLESNIKOV A, SPERL G, et al. iCaRL: Incremental classifier and representation learning[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, USA, 2017: 5533–5542.
    [18] 郭立民, 陈鑫, 陈涛. 基于AlexNet模型的雷达信号调制类型识别[J]. 吉林大学学报:工学版, 2019, 49(3): 1000–1008. doi: 10.13229/j.cnki.jdxbgxb20171056

    GUO Limin, CHEN Xin, and CHEN Tao. Radar signal modulation type recognition based on AlexNet model[J]. Journal of Jilin University:Engineering and Technology Edition, 2019, 49(3): 1000–1008. doi: 10.13229/j.cnki.jdxbgxb20171056
    [19] 秦鑫, 黄洁, 查雄, 等. 基于扩张残差网络的雷达辐射源信号识别[J]. 电子学报, 2020, 48(3): 456–462. doi: 10.3969/j.issn.0372-2112.2020.03.006

    QIN Xin, HUANG Jie, ZHA Xiong, et al. Radar emitter signal recognition based on dilated residual network[J]. Acta Electronica Sinica, 2020, 48(3): 456–462. doi: 10.3969/j.issn.0372-2112.2020.03.006
    [20] KÄDING C, RODNER E, FREYTAG A, et al. Fine-tuning deep neural networks in continuous learning scenarios[C]. The 13th Asian Conference on Computer Vision, Taipei, China, 2016: 588–605.
  • 加载中
图(12) / 表(2)
计量
  • 文章访问数:  320
  • HTML全文浏览量:  129
  • PDF下载量:  105
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-07-12
  • 录用日期:  2022-01-25
  • 修回日期:  2022-01-18
  • 网络出版日期:  2022-02-19
  • 刊出日期:  2022-09-19

目录

    /

    返回文章
    返回