基于动态图神经网络与对比学习的单通道同频盲信号分离方法

罗鸣; 赖金山; 孙鹏钊; 张昆; 张凤荔; 王瑞锦

doi:10.12263/DZXB.20260400

您当前的位置：

首页 >

文章列表页 >

基于动态图神经网络与对比学习的单通道同频盲信号分离方法

射频计算 | 更新时间：2026-10-16

- 基于动态图神经网络与对比学习的单通道同频盲信号分离方法
- Single-Channel Co-Frequency Blind Signal Separation Based on Dynamic Graph Neural Networks and Contrastive Learning
- 电子学报 2026年54卷第6期页码：2493-2508
- 作者机构：
  
  1.电子科技大学，四川成都 610054
  2.成都菁蓉联创科技有限公司，四川成都 610041
- 作者简介：
  
  罗鸣男，2003年3月出生于河南省南阳市。现为电子科技大学信息与软件工程学院硕士研究生。主要研究方向为图神经网络、智能计算、知识图谱。E-mail: 202522090401@std.uestc.edu.cn
  赖金山男，1998年10月出生于四川省资阳市。现为电子科技大学信息与软件工程学院博士研究生。主要研究方向为联邦学习与聚合安全、智能计算、网络与信息安全、区块链。E-mail: 202311090910@std.uestc.edu.cn
  孙鹏钊男，1999年3月出生于河南省平顶山市。现为电子科技大学信息与软件工程学院博士研究生。主要研究方向为图神经网络、推荐系统。E-mail: spz@std.uestc.edu.cn
  张昆男，1982年1月出生于四川省成都市。高级工程师。主要研究方向为国产化基础软硬件。E-mail: zhangkun@cdjrlc.com
  张凤荔女，1963年出生于河南省南阳市。现为电子科技大学教授。主要研究方向为智能计算、联邦学习、网络与信息安全。E-mail: fzhang@uestc.edu.cn
  王瑞锦男，1980年出生于甘肃省天水市。现为电子科技大学教授。主要研究方向为联邦学习与聚合安全、智能计算、区块链、知识图谱、信息安全。E-mail: ruijinwang@uestc.edu.cn
- 基金信息：
  
  国家自然科学基金(62271128);四川省重大科技专项(2025ZDZX0093);四川省科技计划项目(2025ZHRG0006);成都市重点研发支持计划项目(2025-YF08-00128-GX;2025-YF12-00029-RC)
- DOI：10.12263/DZXB.20260400
  中图分类号： TP18;
- 收稿：2026-04-26，
  
  录用：2026-05-19，
  
  纸质出版：2026-06-25
- 稿件说明：
移动端阅览
罗鸣, 赖金山, 孙鹏钊, 等. 基于动态图神经网络与对比学习的单通道同频盲信号分离方法[J]. 电子学报, 2026, 54(06): 2493-2508.

LUO Ming, LAI Jinshan, SUN Pengzhao, et al. Single-Channel Co-Frequency Blind Signal Separation Based on Dynamic Graph Neural Networks and Contrastive Learning[J]. Acta Electronica Sinica, 2026, 54(06): 2493-2508.
罗鸣, 赖金山, 孙鹏钊, 等. 基于动态图神经网络与对比学习的单通道同频盲信号分离方法[J]. 电子学报, 2026, 54(06): 2493-2508. DOI：10.12263/DZXB.20260400

LUO Ming, LAI Jinshan, SUN Pengzhao, et al. Single-Channel Co-Frequency Blind Signal Separation Based on Dynamic Graph Neural Networks and Contrastive Learning[J]. Acta Electronica Sinica, 2026, 54(06): 2493-2508. DOI：10.12263/DZXB.20260400

摘要

在复杂电磁环境下，目标信号与干扰信号在时域、频域高度重叠，单通道接收条件下的同频盲信号分离面临较大挑战。现有基于卷积的分离方法在复杂非平稳干扰条件下，主要存在非局部相关性建模不足、目标信号与干扰信号隐空间解耦能力有限、对未知干扰模式泛化性较弱等问题。本文提出基于动态图神经网络与对比学习的单通道同频盲信号分离方法（Dynamic Graph and Contrastive Learning network，DGCL）。首先，将复基带混合序列映射为I/Q双通道表示，并依托动态图神经网络实现多尺度图时序特征的有效提取，通过多尺度分段、动态构图与图消息传递的协同作用，增强对复杂非平稳干扰的表征能力；其次，设计目标信号匹配驱动的对比学习自编码器，在结构匹配重构的基础上引入对比约束，提升目标信号与干扰信号在隐空间中的可分性；此外，在训练时设计隐空间流形混合增强策略，通过扩展深层特征分布，提升模型对未知干扰模式的鲁棒性。在单载波、多载波及多类干扰条件下的实验表明，DGCL在误码率与均方误差两项指标上均表现出显著优势。相较于常规基准模型WaveNet框架，DGCL的MSE（Mean Square Error）降低了6.7 dB，BER（Bit Error Rate）下降了1.72 × 10

-2

，这些结果验证了本文方法的有效性。

Abstract

In complex electromagnetic environments

target signals and interference signals heavily overlap in both time and frequency domains

posing significant challenges for co-frequency blind signal separation under single-channel reception conditions. Under complex non-stationary interference conditions

existing convolution-based separation methods primarily suffer from insufficient non-local correlation modeling

limited decoupling capability between target and interference signals in the latent space

and weak generalization to unknown interference patterns. To address these issues

this paper proposes a single-channel co-frequency blind signal separation method based on dynamic graph neural networks and contrastive learning (DGCL). First

the complex baseband mixture s

equence is mapped into an in-phase/quadrature (I/Q) dual-channel representation

and dynamic graph neural networks are utilized to effectively extract multi-scale graph temporal features. Through the synergistic effects of multi-scale segmentation

dynamic graph construction

and graph message passing

the representation capability for complex non-stationary interference is enhanced. Second

a contrastive learning autoencoder driven by target signal matching is designed. By introducing contrastive constraints on top of structural matching reconstruction

the separability of the target signal and the interference signal in the latent space is improved. Furthermore

a latent space manifold mixup enhancement strategy is designed during training. By expanding the deep feature distribution

the model’s robustness against unknown interference patterns is elevated. Experiments conducted under single-carrier

multi-carrier

and multi-class interference conditions indicate that DGCL demonstrates significant advantages in both bit error rate and mean square error metrics.Compared with the conventional WaveNet baseline

DGCL reduces MSE (Mean Square Error) by 6.7 dB and decreases BER (Bit Error Rate) by 1.72 × 10

-2

. These results verify the effectiveness of the proposed method.

关键词

Keywords

references

Lancho A , Weiss A , Lee G C F , et al . RF challenge: The data-driven radio frequency signal separation challenge [J ] . IEEE Open Journal of the Communications Society , 2025 , 6 : 4083 - 4100 . DOI: 10.1109/ojcoms.2025.3556319 http://dx.doi.org/10.1109/ojcoms.2025.3556319

Hyvärinen A , Oja E . Independent component analysis: Algorithms and applications [J ] . Neural Networks , 2000 , 13 ( 4/5 ): 411 - 430 . DOI: 10.1016/s0893-6080(00)00026-5 http://dx.doi.org/10.1016/s0893-6080(00)00026-5

Hwang W L , Ho J . Null space component analysis of one-shot single-channel source separation problem [J ] . IEEE Transactions on Signal Processing , 2021 , 69 : 2233 - 2251 . DOI: 10.1109/tsp.2021.3064181 http://dx.doi.org/10.1109/tsp.2021.3064181

Wang Taihui , Yang Feiran , Yang Jun . Convolutive transfer function-based multichannel nonnegative matrix factorization for overdetermined blind source separation [J ] . IEEE/ACM Transactions on Audio , Speech, and Language Processing, 2022 , 30 : 802 - 815 . DOI: 10.1109/taslp.2022.3145304 http://dx.doi.org/10.1109/taslp.2022.3145304

Guo Pengcheng , Yu Miao , Shen Lei , et al . Single-channel blind source separation in wireless communications: A complex-domain deep learning approach [J ] . IEEE Wireless Communications Letters , 2024 , 13 ( 6 ): 1645 - 1649 . DOI: 10.1109/lwc.2024.3384813 http://dx.doi.org/10.1109/lwc.2024.3384813

Su Jie , Jiang Yuting , Ye Yuheng , et al . Patch matter: Dual modality patch contrastive for non-stationary radio signals [J ] . IEEE Transactions on Mobile Computing , 2026 , 25 ( 4 ): 5936 - 5951 . DOI: 10.1109/tmc.2025.3633263 http://dx.doi.org/10.1109/tmc.2025.3633263

王瑞锦 , 王金波 , 张凤荔 , 等 . 联邦原型学习的特征图中毒攻击和双重防御机制 [J ] . 软件学报 , 2025 , 36 ( 3 ): 1355 - 1374 .

Wang Ruijin , Wang Jinbo , Zhang Fengli , et al . Feature map poisoning attack and dual defense mechanism for federated prototype learning [J ] . Journal of Software , 2025 , 36 ( 3 ): 1355 - 1374 . (in Chinese)

Liu Chao , Fu Xue , Wang Yu , et al . Overcoming data limitations: A few-shot specific emitter identification method using self-supervised learning and adversarial augmentation [J ] . IEEE Transactions on Information Forensics and Security , 2024 , 19 : 500 - 513 . DOI: 10.1109/tifs.2023.3324394 http://dx.doi.org/10.1109/tifs.2023.3324394

Yang Jian , Zhu Shaoxian , Wen Zhongyi , et al . Cross-receiver radio frequency fingerprint identification: A source-free adaptation approach [J ] . Sensors , 2025 , 25 ( 14 ): 4451 . DOI: 10.3390/s25144451 http://dx.doi.org/10.3390/s25144451

万升 , 杨健 , 宫辰 . 基于图神经网络的高光谱图像分类研究进展 [J ] . 电子学报 , 2023 , 51 ( 6 ): 1687 - 1709 .

Wan Sheng , Yang Jian , Gong Chen . Advances of hyperspectral image classification based on graph neural networks [J ] . Acta Electronica Sinica , 2023 , 51 ( 6 ): 1687 - 1709 . (in Chinese)

Tian Yu , Alhammadi A , Quran A , et al . A novel approach to wavenet architecture for RF signal separation with learnable dilation and data augmentation [C ] // 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops . Piscataway : IEEE , 2024 : 79 - 80 . DOI: 10.1109/icasspw62465.2024.10627586 http://dx.doi.org/10.1109/icasspw62465.2024.10627586

Henneke L . Improving data-driven RF signal separation with SOI-matched autoencoders [C ] // 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops . Piscataway : IEEE , 2024 : 45 - 46 . DOI: 10.1109/icasspw62465.2024.10626245 http://dx.doi.org/10.1109/icasspw62465.2024.10626245

Damara F , Utkovski Z , Stanczak S . Signal separation in radio spectrum using self-attention mechanism [C ] // 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops . Piscataway : IEEE , 2024 : 99 - 100 . DOI: 10.1109/icasspw62465.2024.10627553 http://dx.doi.org/10.1109/icasspw62465.2024.10627553

Yapar Ç , Jaensch F , Hauffen J C , et al . Demucs for data-driven RF signal denoising [C ] // 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops . Piscataway : IEEE , 2024 : 95 - 96 . DOI: 10.1109/icasspw62465.2024.10627485 http://dx.doi.org/10.1109/icasspw62465.2024.10627485

Naseri M , Fontaine J , Moerman I , et al . A U-Net architecture for time-frequency interference signal separation of RF waveforms [C ] // 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops . Piscataway : IEEE , 2024 : 91 - 92 . DOI: 10.1109/icasspw62465.2024.10626228 http://dx.doi.org/10.1109/icasspw62465.2024.10626228

Zhang Bingxu , Fan Changjun , Liu Shixuan , et al . The expressive power of graph neural networks: A survey [J ] . IEEE Transactions on Knowledge and Data Engineering , 2025 , 37 ( 3 ): 1455 - 1474 . DOI: 10.1109/tkde.2024.3523700 http://dx.doi.org/10.1109/tkde.2024.3523700

Jin Ming , Koh H Y , Wen Qingsong , et al . A survey on graph neural networks for time series: Forecasting, classification, imputation, and anomaly detection [J ] . IEEE Transactions on Pattern Analysis and Machine Intelligence , 2024 , 46 ( 12 ): 10466 - 10485 . DOI: 10.1109/tpami.2024.3443141 http://dx.doi.org/10.1109/tpami.2024.3443141

Shen Yifei , Shi Yuanming , Zhang Jun , et al . Graph neural networks for scalable radio resource management: Architecture design and theoretical analysis [J ] . IEEE Journal on Selected Areas in Communications , 2021 , 39 ( 1 ): 101 - 115 . DOI: 10.1109/jsac.2020.3036965 http://dx.doi.org/10.1109/jsac.2020.3036965

Lee M , Yu Guanding , Dai Huaiyu . Decentralized inference with graph neural networks in wireless communication systems [J ] . IEEE Transactions on Mobile Computing , 2023 , 22 ( 5 ): 2582 - 2598 . DOI: 10.1109/tmc.2021.3125793 http://dx.doi.org/10.1109/tmc.2021.3125793

Yin Xiaoyan , Lin Wanyu , Sun Kexin , et al . A 2 S 2 -GNN: Rigging GNN-based social status by adversarial attacks in signed social networks [J ] . IEEE Transactions on Information Forensics and Security , 2023 , 18 : 206 - 220 . DOI: 10.1109/tifs.2022.3219342 http://dx.doi.org/10.1109/tifs.2022.3219342

Wang Ruijin , Wang Jinbo , Li Xiong , et al . CESA: Communication efficient secure aggregation scheme via sparse graph in federated learning [J ] . Journal of Network and Computer Applications , 2024 , 231 : 103997 . DOI: 10.1016/j.jnca.2024.103997 http://dx.doi.org/10.1016/j.jnca.2024.103997

Huang Wei , Han Andi , Chen Yongqiang , et al . On the comparison between multi-modal and single-modal contrastive learning [C ] // Advances in Neural Information Processing Systems 37 . Neural Information Processing Systems Foundation, Inc. (NeurIPS) , 2024 : 81549 - 81605 . DOI: 10.52202/079017-2592 http://dx.doi.org/10.52202/079017-2592

Yue Zhihan , Wang Yujing , Duan Juanyong , et al . TS2Vec: Towards universal representation of time series [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2022 , 36 ( 8 ): 8980 - 8987 . DOI: 10.1609/aaai.v36i8.20881 http://dx.doi.org/10.1609/aaai.v36i8.20881

Du Mingyang , Pan Jifei , Bi Daping . A contrastive learner for automatic modulation classification [J ] . IEEE Transactions on Wireless Communications , 2025 , 24 ( 4 ): 3575 - 3589 . DOI: 10.1109/twc.2025.3532438 http://dx.doi.org/10.1109/twc.2025.3532438

Zhang Hongyi , Cissé M , Dauphin Y N , et al . Mixup: Beyond empirical risk minimization [C ] // Proceedings of the 6th International Conference on Learning Representations . OpenReview . net , 2018 : 2866 - 2878 .

Verma V , Lamb A , Beckham C , et al . Manifold mixup: Better representations by interpolating hidden states [C/OL ] // Proceedings of the 36th International Conference on Machine Learning . 2019 : 6438 - 6447 . https://dblp.uni-trier.de/db/conf/icml/icml2019.html#conf/icml/VermaLBNMLB19 https://dblp.uni-trier.de/db/conf/icml/icml2019.html#conf/icml/VermaLBNMLB19 .

Hendrycks D , Mu N , Cubuk E D , et al . AugMix: A simple data processing method to improve robustness and uncertainty [C ] // Proceedings of the 8th International Conference on Learning Representations . ICLR , 2020 : 1108 - 1122 .

Mu Shiyuan , Chen Shuai , Zu Yong , et al . Contrastive learning with feature-level augmentation for wireless signal representation [J ] . Electronics , 2025 , 14 ( 13 ): 2728 . DOI: 10.3390/electronics14132728 http://dx.doi.org/10.3390/electronics14132728

He Jialong , Xie Yuelei , Liu Xiangguo . Channel-robust specific emitter identification based on domain-adversarial training of neural networks and multi-feature fusion [J ] . IEEE Access , 2025 , 13 : 153093 - 153104 . DOI: 10.1109/access.2025.3604428 http://dx.doi.org/10.1109/access.2025.3604428

王瑞锦 , 赖金山 , 王金波 . 边缘计算下的联邦学习:安全与隐私挑战 [J ] . 计算 , 2026 , 53 ( 4 ): 24 - 31 .

Wang Ruijin , Lai Jinshan , Wang Jinbo . Federated learning in edge computing: Security and privacy challenges [J ] . Computing Magazine of the CCF , 2026 , 53 ( 4 ): 24 - 31 . (in Chinese)

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

面向中继卫星任务调度的时间约束驱动图强化学习方法

基于层次化掩码图自编码框架的APT威胁检测方法

特征掩码与对比学习融合多维度去过度相关的序列推荐

融合大语言模型与跨域图结构的多模态对话情感识别方法

一种加载介质钉层的宽角低扫描损耗毫米波双极化相控阵