雷达脉内调制识别的改进残差神经网络算法

doi:10.13229/j.cnki.jdxbgxb20200447

摘要/Abstract

摘要：

针对人工提取的特征计算量大且存在主观、不能完全体现信号本质以及生成时频图像耗时过长的问题，提出了一种以改进残差神经网络ResNet32为框架，对雷达时域信号特征进行提取并识别的雷达信号脉内调制的算法。算法建立9种脉内信号的时域信号数据集，输入到ResNet32框架中进行训练、分类、识别。算法节省了大量生成时频图像的时间，并且实验验证算法在低信噪比（SNR）时的识别率更加优秀。在混合信噪比的实验条件中，SNR=-14 dB和SNR=-8 dB时的识别率均达到90%以上。

关键词: 模式识别与智能系统, 雷达信号, 脉内调制识别, 残差神经网络

Abstract:

The artificially extracted features are computationally intensive and subjective, fail to fully reflect the nature of the signal, and take too long to generate time-frequency images. To overcome these problems, We propose an improved residual neural network (ResNet) ResNet32 as a framework to extract and identify radar time-domain signal features. We build a time-domain signal dataset of 9 types of intra-pulse signals and input them into the ResNet32 framework for training and classification. The algorithm saves a lot of time to generate time-frequency images, and the experimental verification algorithm has a better recognition rate at low signal to noise ratio (SNR). In the experimental conditions of mixed SNR, the recognition rate of the 9 modulation types with SNR=-14 dB and SNR=-8 dB achieved more than 90%.

Key words: pattern recognition and intelligent system, radar signal, intra-pulse modulation classification, residual neural network

中图分类号:

TP183

徐卓君,杨雯婷,杨承志,田彦涛,王晓军. 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.

Zhuo-jun XU,Wen-ting YANG,Cheng-zhi YANG,Yan-tao TIAN,Xiao-jun WANG. 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.

图/表 11

表1

相位调制函数"

码型	$? (t)$ 的定义
2PSK	$? (t) = 0 或 π$
4PSK	$? (t) = 0, π / 2, π 或 3 π / 2$
P1码	$? m (n, k) = - π / L M - (2 k - 1) (k - 1) M + n - 1$ $n = 1,2, …, L$ ， $k = 1,2, …, L$
P2码	$? m (n, k) = - π / 2 L (2 n - 1 - L) (2 k - 1 - L)$ $n = 1,2, …, L$ ， $k = 1,2, …, L$ （ $L$ 为偶数）
P3码	$? m (k) = π / M (k - 1) 2$ ， $k = 1,2, …, M$
P4码	$? m (k) = π / M (k - 1) (k - 1 - M)$ ， $k = 1,2, …, M$

表1

图1

表2

表3

图2

表4

图3

图4

表5

表 6

图5

参考文献 15

1	王星, 呙鹏程, 田元荣, 等. 基于BDS-GD的低截获概率雷达信号识别[J]. 北京航空航天大学学报, 2018, 44(3):583-592.
	Wang Xing, Guo Peng-cheng, Tian Yuan-rong, et al. LPI radar signal recognition based on BDS-GD[J]. Journal of Beijing University of Aeronautics and Astronautics, 2018, 44(3):583-592.
2	周志文, 黄高明, 高俊, 等. 一种深度学习的雷达辐射源识别算法[J]. 西安电子科技大学学报:自然科学版, 2017, 44(3):77-82.
	Zhou Zhi-wen, Huang Gao-ming, Gao Jun, et al. Radar emitter identification algorithm based on deep learning[J]. Journal of Xidian University, 2017, 44(3):77-82.
3	Qu Z Y, Mao X J, Deng Z A. Radar signal intra-pulse modulation recognition based on convolutional neural network[J]. IEEE Access, 2018, 6:43874-43884.
4	Qin X, Zha X, Huang J, et al. Radar waveform recognition based on deep residual network[C]∥IEEE 8th Joint International Information Technology and Artificial Intelligence Conference, Chongqing, China, 2019:892-896.
5	胡国兵, 宋军, 李昌利. 低截获概率雷达侦察信号分析及可信性评估[M]. 北京:国防工业出版社, 2017.
6	樊昌信. 通信原理[M]. 北京:国防工业出版社, 2009.
7	张贤达. 通信信号处理[M]. 北京:国防工业出版社, 2000.
8	刘峰, 黄宇, 王泽众, 等. 复杂信号侦察理论及应用[M]. 北京:科学出版社, 2016.
9	王彩云, 何志勇, 宫俊. 基于压缩感知的单脉冲雷达欺骗干扰机研究[J]. 北京航空航天大学学报, 2017, 43(9):1789-1797.
	Wang Cai-yun, He Zhi-yong, Gong Jun. Research on deception jammer against monopulse radar based on compressed sensing[J]. Journal of Beijing University of Aeronautics and Astronautics, 2017, 43(9):1789-1797.
10	He K M, Zhang X Y, Ren S Q, et al. Deep residual learning for image recognition[C]∥IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Seattle,USA, 2016:770-778.
11	Zhang Y B, Sun L L, Yan C G, et al. Adaptive residual networks for high-quality image restoration[J]. IEEE Transactions on Image Processing, 2018, 27(7): 3150-3163.
12	李琳辉, 伦智梅, 连静,等. 基于卷积神经网络的道路车辆检测方法[J]. 吉林大学学报:工学版, 2017,47(2):384-391.
	Li Lin-hui, Zhi-mei Lun, Lian Jing, et al. Convolution neural network-based vehicle detection method[J]. Journal of Jilin University (Engineering and Technology Edition), 2017, 47(2):384-391.
13	Wen Z G, Liu D, Liu X Q, et al. Deep learning based smart radar vision system for object recognition[J]. Journal of Ambient Intelligence and Humanized Computing, 2019,10:829-839.
14	Li B Q, He Y Y. An improved ResNet based on the adjustable shortcut connections[J]. IEEE Access, 2018, 6:18967-18974.
15	郭坚, 漆轩. 基于残差网络的自动调制识别[J]. 计算机工程与设计, 2019, 40(9):2406-2410.
	Guo Jian, Qi Xuan. Automatic modulation classification based on residual network[J]. Computer Engineering and Design, 2019, 40(9):2406-2410.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

参数名称	本文ResNet参数设置
卷积核大小	［3×1］，［5×1］，［3×1］
残差块结构
池化方法	Max Pooling
损失函数（Loss function）	Cross entropy
批量大小（Batch size）	16
训练循环次数（Epoch）	200
学习率	0.001

结构	内容
调制类型	LFM、2FSK、4FSK、2PSK、 4PSK、P1、P2、P3、P4
信号形式	IQ两路信号
样本长度	2048采样点
样本维数
信噪比范围	-20~20 dB，间隔步长为2 dB
样本个数	单信噪比样本个数为1024个

信噪比/dB	本文识别率/%	ResNet18识别率/%
-20	62.22	41.79
-18	63.22	45.64
-16	65.33	47.13
-14	70.33	50.82
-12	74.33	53.89
-10	74.11	54.99
-8	75.22	60.69
-6	78.89	62.69
-4	84.67	64.19
-2	87.33	67.33
0	92.67	69.22
2	95.78	72.44
4	96.33	72.77
6	96.56	74.76
8	97.22	75.74
10	98.00	81.32
12	98.22	85.13
14	98.78	88.52
16	98.44	89.75
18	98.89	92.33
20	98.44	92.57

信噪比/dB	本文识别率/%	ResNet18识别率/%
-20	72.02	43.56
-18	79.00	54.00
-16	87.03	67.11
-14	92.90	75.67
-12	97.59	84.44
-10	99.44	92.78
-8	100.00	97.22
-6	100.00	98.63
-4	100.00	99.67
-2	100.00	100.00
0	100.00	100.00
2	100.00	100.00
4	100.00	100.00
6	100.00	100.00
8	100.00	100.00
10	100.00	100.00
12	100.00	100.00
14	100.00	100.00
16	100.00	100.00
18	100.00	100.00
20	100.00	100.00

[1]	刘富,刘璐,侯涛,刘云. 基于优化MSR的夜间道路图像增强方法[J]. 吉林大学学报(工学版), 2021, 51(1): 323-330.
[2]	刘富, 兰旭腾, 侯涛, 康冰, 刘云, 林彩霞. 基于优化k-mer频率的宏基因组聚类方法[J]. 吉林大学学报(工学版), 2018, 48(5): 1593-1599.