基于注意力机制增强残差网络的雷达信号调制类型识别

doi:10.12382/bgxb.2022.0302

摘要/Abstract

摘要：

针对情报分析中在低信噪比下雷达信号调制样式识别率较低的问题,提出基于注意力机制增强残差网络的雷达调制类型识别算法。利用平滑伪Wigner-Ville分布时频变换的强能量聚集特点将信号调制样式转化为二维时频图像;搭建2层卷积网络和6层残差块的残差网络,并在网络之间穿插加入卷积注意力机制模块,用以增强对特征的关注度,提高特征提取的有效性;将二维时频图像输入到该网络模型中实现调制类型识别。仿真实验结果表明,该算法对6类典型雷达信号调制类型能够有效提取到时频图像的特征,在信噪比0dB以上实现100%的正确率,在信噪比-10dB 下正确率依然能够保持94.2%,具有较强的识别优势。

关键词: 信号调制类型识别, 平滑伪Wigner-Ville分布, 卷积神经网络, 残差网络, 注意力机制

Abstract:

To solve the problem of low recognition rate of radar signal modulation type under the condition of low SNR in intelligence analysis, an radar signal modulation type recognition method based on attention mechanism enhanced residual networks is proposed. Firstly, drawing on the strong energy aggregation characteristics of the smooth pseudo Wigner-Ville distribution time-frequency transform, the signal modulation type is transformed into a two-dimensional time-frequency image. Then, residual networks composed of a two-layer convolution network and a six-layer residual block are built, and the convolutional block attention module is inserted between the networks to enhance the attention to features and improve the effectiveness of feature extraction. Finally, two-dimensional time-frequency images are input into the network model to realize modulation type recognition. The simulation results show that this method can effectively extract the feature of time-frequency images for six typical radar signal modulation types, and can achieve 100% accuracy above 0dB Signal-Noise Ratio, and still maintain 94.2% accuracy under -10dB Signal-Noise Ratio.

Key words: signal modulation type recognition, smooth pseudo Wigner-Ville distribution, convolutional neural network, residual network, attention mechanism

中图分类号:

TN97.1

吴礼洋, 呙鹏程, 刘超, 李文强. 基于注意力机制增强残差网络的雷达信号调制类型识别[J]. 兵工学报, 2023, 44(8): 2310-2318.

WU Liyang, GUO Pengcheng, LIU Chao, LI Wenqiang. Radar Signal Modulation Type Recognition Based on Attention Mechanism Enhanced Residual Networks[J]. Acta Armamentarii, 2023, 44(8): 2310-2318.

图/表 16

图1 SPWVD和WVD时频分布图对比

Fig.1 Comparison between SPWVD and WVD

图2 CBAM结构

Fig.2 Construction of CBAM

图3 CAM结构

Fig.3 Construction of CAM

图4 SAM结构

Fig.4 Construction of SAM

图5 残差块结构

Fig.5 Construction of residual block

图6 基于注意力机制增强的残差网络识别框架

Fig.6 Recognition framework of residual networks enhanced by attention mechanism

表1 各网络层特征参数

Table 1 Parameters of network features

基础层	分支层	输入尺寸× 通道数	输出尺寸× 通道数
	Conv2d_1卷积层	56×56×1	56×56×32
基础网络块1	Conv2d_2卷积层	56×56×32	56×56×64
	CBAM模块	56×56×64	56×56×64
	Residual_1残差块	56×56×64	56×56×64
基础网络块2	Residual_2残差块	56×56×64	28×28×128
	CBAM模块	28×28×128	28×28×128
	Residual_1残差块	28×28×128	28×28×128
基础网络块3	Residual_2残差块	28×28×128	14×14×256
	CBAM模块	14×14×256	14×14×256
	Residual_1残差块	14×14×256	14×14×256
基础网络块4	Residual_2残差块	14×14×256	7×7×512
	CBAM模块	7×7×512	7×7×512
全局平均池化层		7×7×512	1×1×512
线性层		1×1×512	1×1×6

图7 残差块特征参数

Fig.7 Residual block feature parameters

表2 信号参数设置

Table 2 Signal parameter setting

类型	载频/ MHz	采样频率/ MHz	(子)脉宽/ ms	巴克码
CW	7~12	60	1~5
BPSK	7~12	60		7,11,13
LFM	10	60	8~16
FSK	9~12	60	1
NLFM	7~12	60	8~16
QPSK	7~12	60	1

图8 6类调制信号的SPWVD时频分布

Fig.8 Time-frequency distribution of 6 modulation signal types

图9 训练误差、测试误差及识别率

Fig.9 Training loss, test loss and recognition rate

图10 识别率对比

Fig.10 Comparison of recognition rate

图11 输入网络的NLFM信号示例

Fig.11 Example of NLFM

图12 提取到网络层特征可视化图例

Fig.12 Visualization examples of extracted network layer features

图13 不同CBAM数量的识别率对比

Fig.13 Comparison of recognition rate under different numbers of CBAM

表3 不同算法识别率

Table 3 Recognition rate of different algrithms

识别算法	信噪比/dB
识别算法	-20	-10	0
LeNet-FisherDDL^[11]	45.6	67.7	98.0
MRSAF-DBN^[19]	51.4	73.5	98.4
Chrip-Zernike-ResNet^[20]	58.6	70.1	99.1
本文算法	63.1	94.2	100

参考文献 20

[1]	KISHORE T R, RAO K D. Automatic intrapulse modulation classification of advanced LPI radar waveforms[J]. IEEE Transaction on Aerospace Electronic and Systems, 2017, 53(2): 901-904. doi: 10.1109/TAES.2017.2667142 URL
[2]	IGLESIAS V, GRAJAL J, ROYER P, et al. Real-time low-complexity automatic modulation classifier for pulsed radar signals[J]. IEEE Transactions on Aerospace Electronic & Systems, 2015, 51(1): 108-126.
[3]	QUAN D Y, TANG Z Y, WANG X F, et al. LPI Radar Signal Recognition Based on Dual-Channel CNN and Feature Fusion[J]. Symmetry, 2022, 14(3): 570. doi: 10.3390/sym14030570 URL
[4]	ZHOU J L, ZHENG S L, YU X B, et al. Low probability of intercept communication based on structured radio beams using machine learning[J]. IEEE Access, 2019, 7: 169946-169952. doi: 10.1109/Access.6287639 URL
[5]	WAN T, JIANG K L, LIAO J Y, et al. Detection and recognition of LPI radar signals using visibility graphs[J]. Journal of Systems Engineering and Electronics, 2020, 31(6): 1186-1192. doi: 10.23919/JSEE.5971804 URL
[6]	王红卫, 董鹏宇, 陈游, 等. 半监督条件下基于朴素贝叶斯和Choi-Williams时频分布能量积累的雷达信号识别[J]. 电子与信息学报, 2021, 43(3): 589-597.
	WANG H W, DONG P Y, CHEN Y, et al. Recognition method of radar signal based on the energy cumulant of Choi-Williams distribution and improved semi-supervised Naïve Bayes[J]. Journal of Electronics & Information Technology, 2021, 43(3): 589-597. (in Chinese)
[7]	LIU S K, YAN X P, LI P, et al. Radar emitter recognition based on SIFT position and scale features[J]. IEEE Transactions on Circuits and Systems II: Express Briefs, 65(12): 2062-2066. doi: 10.1109/TCSII.2018.2819666 URL
[8]	王星, 王俊迪, 金政芝, 等. 机载雷达告警接收机发展现状及趋势[J]. 雷达学报, 2023, 12(2): 376-388.
	WANG X, WANG J D, JIN Z Z, et al. Current situation and development demands for a radar warning receiver system[J]. Journal of Radars, 2023, 12(2): 376-388. (in Chinese)
[9]	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.
	GUO L M, KOU Y H, CHEN T, 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. (in Chinese)
[10]	XIAO Z L, YAN Z Y. Radar emitter identification based on novel time-frequency spectrum and convolutional neural network[J]. IEEE Communications Letters, 2021, 25(8): 2634-2638. doi: 10.1109/LCOMM.2021.3084043 URL
[11]	呙鹏程, 吴礼洋. 融合卷积特征与判别字典学习的低截获概率雷达信号识别[J]. 兵工学报, 2019, 40(9): 1881-1889. doi: 10.3969/j.issn.1000-1093.2019.09.013
	GUO P C, WU L Y. LPI radar signal recognition with convolution feature and discrimination dictionary learning[J]. Acta Armamentarii, 2019, 40(9): 1881-1889. (in Chinese) doi: 10.3969/j.issn.1000-1093.2019.09.013
[12]	普运伟, 刘涛涛, 郭江, 等. 基于卷积神经网络和模糊函数主脊坐标变换的雷达辐射源信号识别[J]. 兵工学报, 2021, 42(8):1680-1689. doi: 10.3969/j.issn.1000-1093.2021.08.012
	PU Y W, LIU T T, GUO J, et al. Radar emitter signal recognition based on convolutional neural network and coordinate transformation of ambiguity function main ridge[J]. Acta Armamentarii, 2021, 42(8):1680-1689. (in Chinese)
[13]	李东瑾, 杨瑞娟, 董睿杰. 基于深度时频特征学习的雷达辐射源识别[J]. 国防科技大学学报, 2020, 42(6): 112-119.
	LI D J, YANG R J, DONG R J. Radar emitter recognition based on the deep learning of time-frequency feature[J]. Journal of National University of Defense Technology, 2020, 42(6): 112-119. (in Chinese)
[14]	HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, US:IEEEE, 2016: 770-778.
[15]	徐卓君, 杨雯婷, 杨承志, 等. 雷达脉内调制识别的改进残差神经网络算法[J]. 吉林大学学报(工学版), 2021, 51(4): 1454-1460.
	XU Z J, YANG W T, YANG C Z, et al. Improved residual neural network algorithm for radar intra-pulse modulation classification[J]. Journal of Jilin University (Engineering and Technology Edition), 2021, 51(4): 1454-1460. (in Chinese)
[16]	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 X, HUANG J, ZHA X, et al. Radar emitter signal recognition based on dilated residual network[J]. Acta Electronica Sinica, 2020, 48(3): 456-462. (in Chinese) doi: 10.3969/j.issn.0372-2112.2020.03.006
[17]	崔邦彦, 田润澜, 王东风, 等. 基于注意力机制和改进CLDNN的雷达辐射源识别[J]. 系统工程与电子技术, 2021, 43(5): 1224-1231.
	CUI B Y, TIAN R L, WANG D F, et al. Radar emitter identification based on attention mechanism and improved CLDNN[J]. Systems Engineering and Electronics, 2021, 43(5): 1224-1231. (in Chinese)
[18]	FLANDRIN P, BORGNAT P. Time-frequency energy distributions meet compressed sensing[J]. IEEE Transactions on Signal Processing, 2010, 58(6): 2974-2982. doi: 10.1109/TSP.2010.2044839 URL
[19]	董鹏宇, 王红卫, 陈游, 等. 基于模糊函数主脊切片和深度置信网络的雷达辐射源信号识别[J]. 空军工程大学学报(自然科学版), 2020, 21(2): 84-90.
	DONG P Y, WANG H W, CHEN Y, et al. A recognition method of radar emitter signals based on SVD of MRSAF and DBN[J]. Journal of Air Force Engineering University (Natural Science Edition), 2020, 21(2): 84-90. (in Chinese)
[20]	谢存祥, 张立民, 钟兆根. 基于时频特征提取和残差神经网络的雷达信号识别[J]. 系统工程与电子技术, 2021, 43(4): 917-926.
	XIE C X, ZHANG L M, ZHONG Z G. Radar signal recognition based on time-frequency feature extraction and residual neural network[J]. Systems Engineering and Electronics, 2021, 43(4): 917-926. (in Chinese)