基于ResNet-Transformer的通信信号自动调制识别

doi:10.19678/j.issn.1000-3428.0069677

摘要/Abstract

摘要：

自动调制识别(AMR)是通信识别、态势感知和电子侦察等领域的重要环节。由于深度神经网络具有很强的特征提取和分类能力, 使得与传统检测方法相比有着更高的识别精度, 但目前常用的神经网络在提取信号时序信息时存在局限性, 包括高复杂度和低信噪比下识别精度差等问题。针对以上问题, 构建一种基于残差神经网络(ResNet)和Transformer网络(ResNet-Transformer)的决策融合识别方案, 旨在处理更复杂的信噪比情况, 并提高整体的识别准确率。该方案首先通过ResNet的时序记忆特性深度挖掘通信信号的时域特征, 然后结合Transformer网络突出的长距离依赖关系提取能力进一步提升抗噪性能, 最后使用决策融合策略根据每条支路输出得到最终判决结果。实验结果表明, 在开源数据集RML2018.01A上, 该方案在信噪比为10 dB以上时平均识别精度大于93%, 在信噪比为0时仍能保持56%的识别精度, 相比传统网络模型能取得更高的调制识别准确率并且具有良好的抗噪能力。

关键词: 自动调制识别, 通信信号, 残差神经网络, Transformer网络, 决策融合

Abstract:

Automatic Modulation Recognition (AMR) is a crucial component in communication identification, situational awareness, and electronic reconnaissance. Deep neural networks, known for their powerful feature extraction and classification capabilities, offer higher recognition accuracy compared to traditional methods. However, current neural networks exhibit limitations in effectively extracting temporal information from signals, leading to high complexity and poor recognition accuracy under low Signal-to-Noise Ratio (SNR) conditions. To address these issues, this paper proposes a decision fusion recognition scheme based on a Residual Neural Network (ResNet) and Transformer network (ResNet-Transformer). This scheme aims to handle more complex SNR scenarios and improve the overall recognition accuracy. By leveraging the temporal memory characteristics of ResNet to deeply extract time-domain features from communication signals, and combining the outstanding long-distance dependency extraction capabilities of the Transformer network to enhance noise resistance, the proposed scheme employs a decision fusion strategy to obtain the final decision based on the outputs of each branch. Experimental results show that the proposed scheme achieves an average recognition accuracy of over 93% for SNRs above 10 dB and maintains a recognition accuracy of 56% even at an SNR of 0 on the open dataset RML2018.01A. Compared to traditional network models, the proposed scheme achieves a higher modulation recognition accuracy and exhibits a high noise resistance.

Key words: Automatic Modulation Recognition (AMR), communication signal, Residual Neural Network (ResNet), Transformer network, decision fusion

沈丹阳, 麦文. 基于ResNet-Transformer的通信信号自动调制识别[J]. 计算机工程, 2026, 52(5): 383-395.

SHEN Danyang, MAI Wen. Automatic Modulation Recognition of Communication Signals Based on the ResNet-Transformer[J]. Computer Engineering, 2026, 52(5): 383-395.

https://www.ecice06.com/CN/Y2026/V52/I5/383

图/表 19

图1 基于ResNet-Transformer的决策融合识别网络结构

Fig.1 Decision fusion recognition network structure based on ResNet-Transformer

图2 改进的ResNet模型

Fig.2 Improved ResNet model

图3 改进的残差栈结构

Fig.3 Improved residual stack structure

图4 基于Transformer的低SNR识别网络模型

Fig.4 Low SNR recognition network model based on Transformer

图5 调制识别中的Transformer模块

Fig.5 Transformer module in modulation recognition

图6 分类结果融合处理方式

Fig.6 Classification result fusion processing method

图7 24种调制信号在SNR为20 dB时的I/Q时域波形图

Fig.7 I/Q time-domain waveform plots of 24 modulation signals when SNR is 20 dB

图8 在SNR为20 dB和-20 dB下BPSK的时域图、ResNet特征图和Transformer特征图

Fig.8 Time-domain plots, ResNet feature maps, and Transformer feature maps of BPSK when SNR is 20 dB and -20 dB

图9 不同方案在不同SNR条件下的识别率对比

Fig.9 Comparison of recognition rates under different SNR conditions for various schemes

图10 两种模型的学习曲线

Fig.10 Learning curves of two models

图11 ResNet-Transformer识别方案实际训练时间

Fig.11 Actual training time of ResNet-Transformer recognition scheme

图12 不同SNR下本文方案的混淆矩阵结果

Fig.12 Confusion matrix results of the proposed scheme under different SNRs

参考文献 27

1	WEAVER C, COLE C A, KRUMLAND R B, et al. The automatic classification of modulation types by pattern recognition[EB/OL]. [2024-03-08]. https://www.semanticscholar.org/paper/THE-AUTOMATIC-CLASSIFICATION-OF-MODULATION-TYPES-BY-Weaver-Cole/bbfd299d0beb7e66f2662625717327211d352709.
2	LIEDTKE F F . Computer simulation of an automatic classification procedure for digitally modulated communication signals with unknown parameters. Signal Processing, 1984, 6 (4): 311- 323. doi: 10.1016/0165-1684(84)90063-X
3	AZZOUZ E E , NANDI A K . Automatic identification of digital modulation types. Signal Processing, 1995, 47 (1): 55- 69. doi: 10.1016/0165-1684(95)00099-2
4	殷赞, 王超杰, 程子恒, 等. 一种基于注意力机制卷积神经网络模型的自动调制识别算法. 电波科学学报, 2023, 38 (5): 773- 779.
	YIN Z , WANG C J , CHENG Z H , et al. An automatic modulation recognition algorithm based on convolutional neural networks with attention mechanism. Chinese Journal of Radio Science, 2023, 38 (5): 773- 779.
5	O'SHEA T J, CORGAN J, CLANCY T C. Convolutional radio modulation recognition networks[M]//JAYNE C, ILIADIS L. Engineering applications of neural networks. Berlin, Germany: Springer International Publishing, 2016: 213-226.
6	LIU X Y, YANG D Y, EL GAMAL A. Deep neural network architectures for modulation classification[C]//Proceedings of the 51st Asilomar Conference on Signals, Systems, and Computers. Washington D.C., USA: IEEE Press, 2017: 915-919.
7	O'SHEA T J , ROY T , CLANCY T C . Over-the-air deep learning based radio signal classification. IEEE Journal of Selected Topics in Signal Processing, 2018, 12 (1): 168- 179. doi: 10.1109/JSTSP.2018.2797022
8	LIN C , YAN W J , ZHANG L M , et al. A real-time modulation recognition system based on software-defined radio and multi-skip residual neural network. IEEE Access, 2020, 8, 221235- 221245. doi: 10.1109/ACCESS.2020.3043588
9	VASWANI A, SHAZEER N M, PARMAR N, et al. Attention is all you need[EB/OL]. [2024-03-08]. https://arxiv.org/abs/1706.03762.
10	汤泊川, 帕力旦·吐尔逊, 柏洁馨, 等. 结合CNN和Transformer的遥感图像土地覆盖分类方法. 微电子学与计算机, 2024, 41 (4): 64- 73.
	TANG B C , Palidan Tuerxun , BAI J X , et al. Land cover classification method for remote sensing images using CNN and Transformer. Microelectronics & Computer, 2024, 41 (4): 64- 73.
11	WEN Q S, ZHOU T, ZHANG C L, et al. Transformers in time series: a survey[C]//Proceedings of the 32nd International Joint Conference on Artificial Intelligence. [S. l. ]: International Joint Conferences on Artificial Intelligence Organization, 2023: 6778-6786.
12	KONG W S, YANG Q H, JIAO X, et al. A Transformer-based CTDNN structure for automatic modulation recognition[C]//Proceedings of the 7th International Conference on Computer and Communications (ICCC). Washington D.C., USA: IEEE Press, 2021: 159-163.
13	HAMIDI-RAD S, JAIN S. MCformer: a Transformer based deep neural network for automatic modulation classification[C]//Proceedings of the IEEE Global Communications Conference (GLOBECOM). Washington D.C., USA: IEEE Press, 2021: 1-6.
14	SU H, FAN X Y, LIU H J. Robust and efficient modulation recognition with pyramid signal Transformer[C]//Proceedings of the 2022 IEEE Global Communications Conference. Washington D.C., USA: IEEE Press, 2022: 1868-1874.
15	杨静雅, 齐彦丽, 周一青, 等. CNN-Transformer轻量级智能调制识别算法. 西安电子科技大学学报, 2023, 50 (3): 40- 49.
	YANG J Y , QI Y L , ZHOU Y Q , et al. Algorithm for recognition of lightweight intelligent modulation based on the CNN-Transformer networks. Journal of Xidian University, 2023, 50 (3): 40- 49.
16	牛瑞婷, 严天峰, 高锐, 等. 低信噪比下基于深度学习TCNN-MobileNet的调制识别. 计算机工程, 2024, 50 (7): 204- 215. doi: 10.19678/j.issn.1000-3428.0068243
	NUI R T , YAN T F , GAO R , et al. Deep learning TCNN-MobileNet-based modulation recognition under low signal-to-noise radio. Computer Engineering, 2024, 50 (7): 204- 215. doi: 10.19678/j.issn.1000-3428.0068243
17	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C., USA: IEEE Press, 2016: 770-778.
18	IOFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on Machine Learning. [S. l. ]: PMLR, 2015: 448-456.
19	刘文婷, 卢新明. 基于计算机视觉的Transformer研究进展. 计算机工程与应用, 2022, 58 (6): 1- 16.
	LIU W T , LU X M . Research progress of transformer based on computer vision. Computer Engineering and Applications, 2022, 58 (6): 1- 16.
20	邱大伟, 刘子辰, 周一青, 等. 基于Transformer神经网络的滚动轴承故障类型识别. 高技术通讯, 2021, 31 (1): 1- 11.
	QIU D W , LIU Z C , ZHOU Y Q , et al. A novel fault type detection method of rolling bearing using Transformer neural networks. Chinese High Technology Letters, 2021, 31 (1): 1- 11.
21	赵雄文, 郭春霞, 李景春. 基于高阶累积量和循环谱的信号调制方式混合识别算法. 电子与信息学报, 2016, 38 (3): 674- 680.
	ZHAO X W , GUO C X , LI J C . Mixed recognition algorithm for signal modulation schemes by high-order cumulants and cyclic spectrum. Journal of Electronics & Information Technology, 2016, 38 (3): 674- 680.
22	谭晓衡, 褚国星, 张雪静, 等. 基于高阶累积量和小波变换的调制识别算法. 系统工程与电子技术, 2018, 40 (1): 171- 177.
	TAN X H , CHU G X , ZHANG X J , et al. Modulation recognition algorithm based on high-order cumulants and wavelet transform. Systems Engineering and Electronics, 2018, 40 (1): 171- 177.
23	梁先明, 倪帆, 陈文洁, 等. 基于时频Grad-CAM的调制识别网络可解释分析. 西南交通大学学报, 2024, 59 (5): 1215- 1224.
	LIANG X M , NI F , CHEN W J , et al. Interpretable analysis of modulation recognition networks based on time-frequency Grad-CAM. Journal of Southwest Jiaotong University, 2024, 59 (5): 1215- 1224.
24	郭帅, 陈婷, 王鹏辉, 等. 基于角度引导Transformer融合网络的多站协同目标识别方法. 雷达学报, 2023, 12 (3): 516- 528.
	GUO S , CHEN T , WANG P H , et al. Multistation cooperative radar target recognition based on an angle-guided Transformer fusion network. Journal of Radars, 2023, 12 (3): 516- 528.
25	邹涵, 张天骐, 马焜然, 等. 基于多特征融合的MIMO-OFDM系统单-混信号调制识别算法. 系统工程与电子技术, 2024, 46 (4): 1456- 1465.
	ZOU H , ZHANG T Q , MA K R , et al. Single-mixed signal modulation and recognition algorithm for MIMO-OFDM system based on multi-feature fusion. Systems Engineering and Electronics, 2024, 46 (4): 1456- 1465.
26	WEST N E, O'SHEA T. Deep architectures for modulation recognition[C]// Proceedings of the IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN). Washington D.C., USA: IEEE Press, 2017: 1-6.
27	金建国. 基于深度神经网络的调制信号识别算法研究[D]. 杭州: 杭州电子科技大学, 2023.
	JIN J G. Research on modulation signal recognition algorithm based on deep neural network[D]. Hangzhou: Hangzhou Dianzi University, 2023. (in Chinese)

[1]	王沙沙, 李帷韬, 刘星宇, 高辉. 基于层级注意力的域自适应遥感图像分割[J]. 计算机工程, 2026, 52(4): 176-186.
[2]	陈东吉, 赖惠成, 高古学, 马骏, 李俊凯, 权虎拓. 知识蒸馏Transformer的人物交互检测[J]. 计算机工程, 2026, 52(1): 206-216.
[3]	钱存元, 陈国强, 李柱培. 基于机器学习的牵引逆变器IGBT间歇开路故障诊断方法[J]. 计算机工程, 2026, 52(1): 369-380.
[4]	张天森, 徐晓娜, 赵悦, 张新宁. 基于级联Transformer和U-Net的MRI肝脏图像分割[J]. 计算机工程, 2025, 51(10): 308-318.
[5]	崔允贺, 赵建朋, 杨鸿珍, 李显超. 基于曲线决策融合的SDN饱和攻击检测方法[J]. 计算机工程, 2024, 50(8): 123-132.
[6]	张池, 王忠, 姜添豪, 谢康民. 基于并行多注意力的语音增强网络[J]. 计算机工程, 2024, 50(4): 68-77.
[7]	王思成, 江浩, 陈晓. 具有跨尺度Transformer的高效多视图立体网络[J]. 计算机工程, 2024, 50(11): 266-275.
[8]	张炯, 王丽芳, 蔺素珍, 秦品乐, 米嘉, 刘阳. 局部全局特征耦合与交叉尺度注意的医学图像融合[J]. 计算机工程, 2023, 49(3): 238-247.
[9]	田炜, 雷志超, 王楚正. 基于多期相注意力融合网络的肝脏病灶CT影像分类研究[J]. 计算机工程, 2023, 49(2): 231-237.
[10]	彭桐歆, 韩勇, 王程, 张志浩. 面向短时地铁客流量预测的混合深度学习模型[J]. 计算机工程, 2022, 48(5): 297-305.
[11]	贺煜航, 刘棪, 陈刚. 基于自适应图卷积网络的心电图多标签分类模型[J]. 计算机工程, 2022, 48(12): 261-269.
[12]	王静，朱翠涛. 重加权分布式多目标决策融合算法[J]. 计算机工程, 2013, 39(11): 87-90.
[13]	许一菲, 肖俊, 武和雷. 基于识别信心决策融合的分块PCA方法[J]. 计算机工程, 2012, 38(10): 148-150.
[14]	张志民, 欧建平, 皇甫堪. 数字通信信号调制自动识别研究[J]. 计算机工程, 2011, 37(6): 153-156.
[15]	张龙飞, 张跃. 一种多导联QRS波实时检测算法[J]. 计算机工程, 2011, 37(16): 282-284.

选择文件类型/文献管理软件名称

选择包含的内容