基于深层残差网络和三元组损失的雷达信号识别方法

doi:10.3969/j.issn.1001-506X.2020.11.12

Abstract

Abstract:

To solve the problem that the classification network is difficult to effectively expand the number of classifications, a radar signal recognition method based on deep residual network and triplet loss is proposed. This method firstly takes the radar signal as the input of the deep residual network, maps the radar signal to 128-dimensional Euclidean space through one-dimensional convolution, and obtains the signal's eigenvector; then uses the triplet loss function to adjust the network parameters so that the Euclidean distance of feature vectors between homogeneous signals decreases and the distance between different types of signals increases; finally, the classification of the signals is realized through a sample library-based recognition algorithm. Experimental results show that compared with traditional classification networks, this method ensures the accuracy of recognition while enabling the model to effectively expand the number of classifications.

Key words: radar signal recognition, deep residual network, triplet loss function, one-dimensional convolution

CLC Number:

TN974

Limeng SHI, Chengzhi YANG, Hongchao WU. Radar signal recognition method based on deep residual network and triplet loss[J]. Systems Engineering and Electronics, 2020, 42(11): 2506-2512.

Figures/Tables 15

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Table 1

Table 2

Table 3

Fig.6

Table 4

Fig.7

Table 5

Fig.8

Fig.9

Fig.10

References 17

1	SALAKHUTDINOV R, HINTON G E. Deep Boltzmann machines[C]//Proc.of the 12th International Intelligence and Statistics, 2009: 448-455.
2	BOASHASH B , BLACK P J . An efficient real-time implementation of the Wigner-Ville distribution[J]. IEEE Trans.on Acoustics Speech and Signal Processing, 1987, 35 (11): 1611- 1618. doi: 10.1109/TASSP.1987.1165070
3	DELPRAT N , ESCUDIE B . Asymptotic wavelet and Gabor analysis: extraction of instantaneous frequencies[J]. IEEE Trans.on Information Theory, 1992, 38 (2): 644- 664. doi: 10.1109/18.119728
4	HE M, MAO Y, HAN J. A method of extracting radar in-pulse Characteristics in Low SNR[EB/OL].[2020-02-19]. https://ieeexplore.ieee.org/document/4129693.
5	许程成, 周青松, 张剑云, 等. 导数约束平滑条件下基于模糊函数特征的雷达辐射源信号识别方法[J]. 电子学报, 2018, 46 (7): 1663- 1668.
	XU C C , ZHOU Q S , ZHANG J Y , et al. Radar emitter recognition based on ambiguity function features with derivative constraint on smoothing[J]. Acta Electronica Sinica, 2018, 46 (7): 1663- 1668.
6	胡国兵, 徐立中, 徐淑芳, 等. 基于能量聚焦效率检验的信号脉内调制识别[J]. 通信学报, 2013, 34 (6): 136- 145.
	HU G B , XU L Z , XU S F , et al. Intrapulse modulation recognition of signals based on statistical test of energy focusing efficiency[J]. Journal on Communications, 2013, 34 (6): 136- 145.
7	徐雄. 采用改进型AlexNet的辐射源目标个体识别方法[J]. 电讯技术, 2018, 58 (6): 625- 630.
	XU X . Radiation source target individual recognition based on improved AlexNet[J]. Telecommunications Engineering, 2018, 58 (6): 625- 630.
8	LIU N B , XU Y N , TIAN Y H , et al. Background classification method based on deep learning for intelligent automotive radar target detection[J]. Future Generation Computer Systems, 2019, 94 (5): 524- 535.
9	刘赢, 田润澜, 王晓峰. 基于深层卷积神经网络和双谱特征的雷达信号识别方法[J]. 系统工程与电子技术, 2019, 41 (9): 1998- 2005.
	LIU Y , TIAN R L , WANG X F . Radar signal recognition method based on deep convolutional neural network and bispectrum feature[J]. Systems Engineering and Electronics, 2019, 41 (9): 1998- 2005.
10	O'SHEA T J , ROY T , CLANCY T C . Over the air deep learning based radio signal classifica-tion[J]. IEEE Journal of Selected Topics in Signal Processing, 2018, 12 (1): 168- 179. doi: 10.1109/JSTSP.2018.2797022
11	HEATON J I G , YOSHUA B , AARON C . Deep learning[J]. Genetic Programming and Evolvable Machines, 2018, 19 (1-2): 305- 307.
12	郭晨, 简涛, 徐从安, 等. 基于深度多尺度一维卷积神经网络的雷达舰船目标识别[J]. 电子与信息学报, 2019, 41 (6): 1302- 1309.
	GUO C , JIAN T , XUN C A , et al. Radar HRRP target recognition based on deep multi-scale 1D convolutional neural network[J]. Journal of Electronics and Information Technology, 2019, 41 (6): 1302- 1309.
13	WANG S H , PHILLIPS P , SUI Y , et al. Classification of alzheimer's disease based on eight-layer convolutional neural network with leaky rectified linear unit and max pooling[J]. Journal of Medical Systems, 2018.
14	HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]// Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
15	SCHROFF F, KALENICHENKO D, PHILBIN J. FaceNet: a unified embedding for face recognition and clustering[C]//Proc.of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018: 815-823.
16	BRIDLE J S. Probabilistic interpretation of feedforward classification network outputs, with relationships to statistical pattern recognition[M]//Neurocomputing: Algorithms, Architectures and Applications, Berlin Heidelberg: Springer, 1990.
17	HE K, ZHANG X, REN S, et al. Delving deep into rectifiers: surpassing human-level per-formance on imageNet classification[C]//Proc. of the IEEE International Conference on Computer Vision, 2015: 1026-1034.

网络结构	激活函数	输出维度
输入层	无	1 024
归一化层1	无	1 024
一维残差堆栈1	ReLU	32×512
一维残差堆栈2	ReLU	32×256
一维残差堆栈3	ReLU	32×128
一维残差堆栈4	ReLU	32×64
一维残差堆栈5	ReLU	32×32
一维残差堆栈6	ReLU	32×16
平坦层	无	512
全连接层1	ReLU	256
全连接层2	ReLU	128
归一化层2(输出层)	无	128

标签	雷达序号	调制方式	载频/MHz	采样率/MHz	脉宽/μs	编码方式	信噪比/dB	每信噪比下的数量
0	雷达1	CW	300	2 000	0.5	无	-20~10	1 000
1	雷达2	CW	500	2 000	0.3	无	-20~10	1 000
2	雷达3	LFM	400~430	2 000	1.2	无	-20~10	1 000
3	雷达4	LFM	600~650	2 000	0.8	无	-20~10	1 000
4	雷达5	BPSK	300	2 000	0.5	13位巴克码	-20~10	1 000
5	雷达6	BFSK	250、300	2 000	0.7	13位巴克码	-20~10	1 000

参数	数值(方法)
epoch	1 000
batchsize	64
学习率	1e-3
优化器	Adam
距离阈值α	0.2

k值	信噪比/dB
k值	-10	-8	-6	-4	-2
1	0.250	0.436	0.583	0.773	0.892
2	0.249	0.439	0.582	0.778	0.893
3	0.251	0.441	0.588	0.780	0.893
4	0.252	0.446	0.593	0.783	0.894
5	0.254	0.453	0.612	0.784	0.895
＞5	0.254	0.453	0.612	0.784	0.895

标签	雷达序号	调制方式	载频/MHz	采样率/MHz	脉宽/μs	编码方式	信噪比/dB	每信噪比下的数量
6	雷达7	QPSK	400	2 000	0.5	格雷码	-20~10	1 000
7	雷达8	CW	100	2 000	1.0	无	-20~10	1 000
8	雷达9	LFM	300~350	2 000	1.5	无	-20~10	1 000
9	雷达10	BPSK	200	2 000	0.3	13位巴克码	-20~10	1 000
10	雷达11	CW	350	2 000	0.8	无	-20~10	1 000

Radar signal recognition method based on deep residual network and triplet loss

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 15

References 17

Related Articles 7

Recommended Articles

Metrics

Comments

[1]	Zhaoguo HOU, Huawei WANG, Liang ZHOU, Qiang FU. Fault diagnosis of rotating machinery based on improved deep residual network [J]. Systems Engineering and Electronics, 2022, 44(6): 2051-2059.
[2]	Pengyu CAO, Chengzhi YANG, Limeng SHI, Hongchao WU. Unknown radar signal processing based on PSO-DBSCAN and SCGAN [J]. Systems Engineering and Electronics, 2022, 44(4): 1158-1165.
[3]	Bangyan CUI, Runlan TIAN, Dongfeng WANG, Gang CUI, Jingyuan SHI. Radar emitter identification based on attention mechanism and improved CLDNN [J]. Systems Engineering and Electronics, 2021, 43(5): 1224-1231.
[4]	Yixiao ZHANG, Wenpu GUO, Kai KANG, Yunlong YAO, Pan WANG. Key radar signal fast recognition method based on clustering and time-series correlation [J]. Systems Engineering and Electronics, 2020, 42(3): 597-602.
[5]	Jiadong YI, Jie YANG. Radar signal recognition based on IFOA-SA-BP neural network [J]. Systems Engineering and Electronics, 2020, 42(12): 2735-2741.
[6]	LIU Ying, TIAN Runlan, WANG Xiaofeng. Radar signal recognition method based on deep convolutional neural network and bispectrum feature [J]. Systems Engineering and Electronics, 2019, 41(9): 1998-2005.
[7]	QU Zhiyu, MAO Xiaojie, HOU Changbo. Radar signal recognition based on singular value entropy and fractal dimension [J]. Systems Engineering and Electronics, 2018, 40(2): 303-307.