基于二值神经网络的辐射源信号识别方法

doi:10.12305/j.issn.1001-506X.2024.11.33

系统工程与电子技术 ›› 2024, Vol. 46 ›› Issue (11): 3912-3919.doi: 10.12305/j.issn.1001-506X.2024.11.33

基于二值神经网络的辐射源信号识别方法

王慧赋¹^,², 梅明飞¹^,², 齐亮², 柴恒³, 陶诗飞¹^,*

1. 南京理工大学电子工程与光电技术学院, 江苏南京 210094
2. 南湖实验室, 浙江嘉兴 314000
3. 中国船舶集团有限公司第七二三研究所, 江苏扬州 225000

收稿日期:2023-07-31 出版日期:2024-10-28 发布日期:2024-11-30
通讯作者: 陶诗飞
作者简介:王慧赋(1998—), 男, 硕士研究生, 主要研究方向为辐射源信号识别、深度学习
梅明飞(2000—), 男, 硕士研究生, 主要研究方向为雷达目标识别、深度学习
齐亮(1981—), 男, 高级工程师, 硕士, 主要研究方向为无线通信技术、信号智能识别技术
柴恒(1982—), 男, 高级工程师, 硕士, 主要研究方向为电子侦察设计
陶诗飞(1987—), 男, 副教授, 硕士, 主要研究方向为雷达侦察、目标识别、电磁隐身技术
基金资助:
海洋防务基金资助课题

Radiation source signal recognition method based on binary neural networks

Huifu WANG¹^,², Mingfei MEI¹^,², Liang QI², Heng CHAI³, Shifei TAO¹^,*

1. School of Electronic and Optical Engineering, Nanjing University of Science and Technology, Nanjing 210094, China
2. Nanhu Laboratory, Jiaxing 314000, China
3. 723 Research Institute, Shipbuilding Group Limited of China, Yangzhou 225000, China

Received:2023-07-31 Online:2024-10-28 Published:2024-11-30
Contact: Shifei TAO

摘要/Abstract

摘要：

针对用于辐射源信号识别的神经网络存在参数冗余、运算量庞大等问题, 提出一种基于二值神经网络的辐射源信号识别方法。该方法指出利用卷积层效用值衡量神经网络卷积层的重要性, 根据卷积层效用值的大小, 将重要的卷积层保留为实值, 其余卷积层进行二值化处理。实验结果表明, 在信噪比大于-9 dB时, 采用该方法得到的二值神经网络的信号识别准确率相比于实值卷积神经网络降低了0.5%, 而网络参数内存大小降低了83.4%, 网络运算次数降低了83.8%, 网络运算复杂度降低了85.8%, 易于部署在各种硬件平台上。

关键词: 辐射源信号识别, 二值神经网络, 卷积层效用值, 网络复杂度

Abstract:

In response to the issues of parameter redundancy and high computational complexity in neural networks used for radiation source signal recognition, a radiation source signal recognition method based on binary neural network is proposed. The method proposes using the utility value of the convolution layer to measure the importance of the neural network's convolution layers. Based on the size of the utility value of the convolution layer, important convolution layers are retained as real values, while the remaining convolution layers are binarized. The experimental results show that when the signal-to-noise ratio is greater than -9 dB, the accuracy of signal recognition of the binary neural network obtained using this method is reduced by 0.5% compared to the real-valued convolutional neural network, while the network parameter memory size is reduced by 83.4%, the network computation is reduced by 83.8%, and the network computing complexity is reduced by 85.8%, and it is easy to deploy on various hardware platforms.

Key words: radiation source signal recognition, binary neural network, convolution layer utility value, network complexity

中图分类号:

TN957.51

王慧赋, 梅明飞, 齐亮, 柴恒, 陶诗飞. 基于二值神经网络的辐射源信号识别方法[J]. 系统工程与电子技术, 2024, 46(11): 3912-3919.

Huifu WANG, Mingfei MEI, Liang QI, Heng CHAI, Shifei TAO. Radiation source signal recognition method based on binary neural networks[J]. Systems Engineering and Electronics, 2024, 46(11): 3912-3919.

图/表 12

图1

图2

图3

图4

图5

表1

表2

表3

表4

图6

表5

表6

参考文献 31

1	YANG H H , YANG X P , LIU S K , et al. Radar emitter multi-label recognition based on residual network[J]. Defence Technology, 2022, 18 (3): 410- 417. doi: 10.1016/j.dt.2021.02.005
2	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. doi: 10.3969/j.issn.1001-506X.2019.09.12
	LIU Y , TIAN R L , WANG X F . Radar signal recognition method based on deep convolutional neural network and bispectral features[J]. Systems Engineering and Electronics, 2019, 41 (9): 1998- 2005. doi: 10.3969/j.issn.1001-506X.2019.09.12
3	MUHAMMAD Y S , WANG J H , XV M M , et al. Ship detection based on deep learning using SAR imagery: a systematic literature review[J]. Soft Computing, 2023, 27 (1): 63- 84. doi: 10.1007/s00500-022-07522-w
4	YAO Y , WANG Z H . Radar signal recognition based on time-frequency and CNN preprocessing[J]. Journal of Detection and Control, 2018, 40 (6): 99- 105.
5	SHI L M , YANG C Z . Recognition method of radar signal mo-dulation method based on deep network[J]. Journal of Ordnance Equipment Engineering, 2021, 42 (6): 190-193, 218. doi: 10.11809/bqzbgcxb2021.06.033
6	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. doi: 10.12305/j.issn.1001-506X.2021.05.09
	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. doi: 10.12305/j.issn.1001-506X.2021.05.09
7	GAO J P , WANG X , WU R W , et al. A new modulation recognition method based on flying fish swarm algorithm[J]. IEEE Access, 2021, 9, 76689- 76706. doi: 10.1109/ACCESS.2021.3079131
8	TAO S F , XIAO S , GONG S , et al. Recognition of electromagnetic signals based on the spiking convolutional neural network[J]. Wireless Communications and Mobile Computing, 2022, 2022, 2395996.
9	MUHAMMAD U, LEE J A. AMC-IoT: automatic modulation classification using efficient convolutional neural networks for low powered IoT devices[C]//Proc. of the International Confe-rence on Information and Communication Technology Convergence, 2020: 288-293.
10	WANG T Y, JIN Y H. Modulation recognition based on lightweight neural networks[C]//Proc. of the 13th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics, 2020: 468-472.
11	ZHANG F X , LUO C B , XV J L , et al. An efficient deep learning model for automatic modulation recognition based on parameter estimation and transformation[J]. IEEE Communications Letters, 2021, 25 (10): 3287- 3290. doi: 10.1109/LCOMM.2021.3102656
12	WANG N , LIU Y X , MA L , et al. Multidimensional CNN-LSTM network for automatic modulation classification[J]. Electronics, 2021, 10 (14): 1649. doi: 10.3390/electronics10141649
13	权宇, 李志欣, 张灿龙, 等. 融合深度扩张网络和轻量化网络的目标检测模型[J]. 电子学报, 2020, 48 (2): 390- 397.
	QUAN Y , LI Z X , ZHANG C L , et al. Fusing deep dilated convolutions network and light-weight network for object detection[J]. ACTA Electonica Sinica, 2020, 48 (2): 390- 397.
14	WANG Z W, ZHENG Q, LU J W, et al. Deep hashing with active pairwise supervision[C]//Proc. of the European Confe-rence on Computer Vision, 2020: 522-538.
15	WANG Z W, XIAO H, LU J W, et al. Generalizable mixed-precision quantization via attribution rank preservation[C]//Proc. of the IEEE/CVF International Conference on Computer Vision, 2021: 2-9.
16	GONG R H, LIU X L, JIANG S H, et al. Differentiable soft quantization: bridging full-precision and low-bit neural networks[C]//Proc. of the IEEE/CVF International Conference on Computer Vision, 2019: 4852-4861.
17	YANG J W, SHEN X, XING J, et al. Quantization networks[C]// Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 7308-7316.
18	ULLRICH K, MEEDS E, WELLING M, et al. Soft weight sharing for neural network compression[C]//Proc. of the International Conference on Learning Representations, 2017.
19	ZHANG D Q, YANG J L, YE D Q, et al. LQ-nets: learned quantization for highly accurate and compact deep neural networks[C]//Proc. of the European Conference on Computer Vision, 2018: 365-382.
20	QIN H T , GONG R H , LIU X L , et al. Binary neural networks: a survey[J]. Pattern Recognition, 2020, 105, 1737- 1746.
21	CHEN T L, ZHANG Z Y, OUYANG X, et al. "BNN-BN=?": training binary neural networks without batch normalization[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 4619-4629.
22	SHEN Z Q, LIU Z C, QIN J, et al. S2-BNN: bridging the gap between self-supervised real and 1-bit neural networks via guided distribution calibration[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 2165-2174.
23	XUE P , LU Y , CHANG J F , et al. Self-distribution binary neural networks[J]. Applied Intelligence, 2022, 52 (12): 13870- 13882. doi: 10.1007/s10489-022-03348-z
24	WANG Z W , LU J W , WU Z Y , et al. Learning efficient binarized object detectors with information compression[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2022, 44 (6): 3082- 3095. doi: 10.1109/TPAMI.2021.3050464
25	JI H, XU W Y, GAN L, et al. Modulation recognition based on lightweight residual network via binary quantization[C]//Proc. of the 7th International Conference on Computer and Communications, 2021: 154-158.
26	RASTEGARI M, ORDONEZ V, REDMON J, et al. Xnor-net: imagenet classification using binary convolutional neural networks[C]//Proc. of the European Conference on Computer Vision, 2016: 525-542.
27	HUANG D K , YAN X P , HAO X H , et al. Low SNR multi-emitter signal sorting and recognition method based on low-order cyclic statistics CWD time-frequency images and the YOLOv5 deep learning model[J]. Sensors, 2022, 22 (20): 7783. doi: 10.3390/s22207783
28	肖帅, 龚帅阁, 李想, 等. FPGA平台轻量化卷积神经网络辐射源信号识别方法[J]. 计算技术与自动化, 2023, 42 (4): 140- 146.
	XIAO S , GONG S G , LI X , et al. Emitter signal identification method with lightweight CNN on FPGA platform[J]. Computing Technology and Automation, 2023, 42 (4): 140- 146.
29	PAVLO M, STEPHEN T, TERO K, et al. Pruning convolutional neural networks for resource efficient inference[EB/OL]. [2023-05-31]. https://arxiv.org/abs/1611.06440.
30	BOYD S , PARIKH N , CHU E , et al. Distributed optimization and statistical learning via the alternating direction method of multipliers[J]. Foundations and Trends in Machine Learning, 2011, 3 (1): 1- 122.
31	王子为, 鲁继文, 周杰. 基于自适应梯度优化的二值神经网络[J]. 电子学报, 2023, 51 (2): 257- 266.
	WANG Z W , LU J W , ZHOU J . Learning adaptive gradients for binary neural networks[J]. ACTA Electonica Sinica, 2023, 51 (2): 257- 266.

信号类型	参数	取值范围
CW	载频f_c	[1/8, 3/8]f_s
LFM	起始频率f₀	[1/16, 3/16]f_s
EQFM	带宽B	[1/16, 3/16]f_s
SFM	起始频率f₀	[1/16, 3/16]f_s
SFM	带宽B	[1/20, 1/10]f_s
COSTAS	基准频率f₀	[1/40, 1/4]f_s
COSTAS	跳频序列N_F	{4, 5, 6}
BPSK	载频f_c	[1/16, 3/16]f_s
BPSK	巴克码长度	{5, 7, 11, 13}
LFM_FSK	基准频率f₀	[1/40, 1/4]f_s
LFM_FSK	子码带宽B	[1/40, 1/20]f_s
LFM_SFM	起始频率f₀	[1/16, 3/16]f_s
LFM_SFM	带宽B	[1/16, 3/16]f_s

超参数	取值
训练轮次	40
批次大小	64
优化器	SGD
学习率	0.01
权重衰减	0.005
动量	0.9

卷积层	效用值
Conv1	0.168 6
Conv2	0.134 9
Conv3	0.150 7
Conv4	0.156 7
Conv5	0.199 6
Conv6	0.189 4

方法	信噪比/dB
方法	-9	-6	-3	0	3
CNN	66.1	93.6	99.7	99.7	99.9
文献[26]	61.1	90.1	98.8	99.6	99.7
本文方法	65.2	93.1	99.1	99.7	99.7

方法	信噪比/dB
方法	-8	-5	-2	1	4
CNN	76.1	96.3	99.7	99.9	99.9
文献[26]	71.6	93.4	99.3	99.7	99.8
本文方法	75.3	96.2	99.3	99.8	99.9

基于二值神经网络的辐射源信号识别方法

Radiation source signal recognition method based on binary neural networks

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 31

相关文章 7

编辑推荐

Metrics

本文评价

网络	参数内存/KB	运算次数/MACs	运算复杂度/FLOPs
CNN	2 156.2	521.5 M	260.8 M
BNN	357.1	84.04 M	37.1 M

[1]	聂千祁, 沙明辉, 朱应申. 基于改进残差神经网络的雷达信号识别方法[J]. 系统工程与电子技术, 2024, 46(10): 3356-3364.
[2]	徐桂光, 王旭东, 汪飞, 胡国兵, 高涌荇, 罗泽虎. 基于SE-ResNeXt网络的低信噪比LPI雷达辐射源信号识别[J]. 系统工程与电子技术, 2022, 44(12): 3676-3684.
[3]	高诗飏, 董会旭, 田润澜, 张歆东. 基于SRNN+Attention+CNN的雷达辐射源信号识别方法[J]. 系统工程与电子技术, 2021, 43(12): 3502-3509.
[4]	孙艺聪, 田润澜, 王晓峰, 董会旭, 戴普. 基于改进CLDNN的辐射源信号识别[J]. 系统工程与电子技术, 2021, 43(1): 42-47.
[5]	李亚兰, 金炜东, 葛鹏. 基于VMD和特征融合的辐射源信号识别[J]. 系统工程与电子技术, 2020, 42(7): 1499-1503.
[6]	黄颖坤, 金炜东, 余志斌, 吴昀璞. 基于深度学习和集成学习的辐射源信号识别[J]. 系统工程与电子技术, 2018, 40(11): 2420-.
[7]	李序,张葛祥,荣海娜. 基于加权K-近邻法和SVC的雷达辐射源信号识别[J]. Journal of Systems Engineering and Electronics, 2010, 32(6): 1215-1219.