面向隐私安全与信道偏移的轻量化短波信号调制识别方法

姚怡舟; 邓文; 李保国

doi:10.11999/JEIT251017

面向隐私安全与信道偏移的轻量化短波信号调制识别方法

doi: 10.11999/JEIT251017 cstr: 32379.14.JEIT251017

1.
武警部队高岭训练基地北京 101507
2.
中国人民解放军63892部队洛阳 471003
3.
国防科技大学电子科学学院长沙 410073

详细信息

作者简介:
姚怡舟：男，工程师，研究方向为信号调制识别

邓文：男，助理研究员，研究方向为智能电子对抗与评估技术

李保国：男，副教授，研究方向为智能电子侦察技术

通讯作者:
邓文　dw0907@nudt.edu.cn

中图分类号: TN971.1
计量
- 文章访问数: 18
- HTML全文浏览量: 9
- PDF下载量: 0
- 被引次数: 0
出版历程
- 修回日期: 2026-01-22
- 录用日期: 2026-01-22
- 网络出版日期: 2026-02-11

Towards Privacy-Preserving and Lightweight Modulation Recognition for Short-Wave Signals under Channel Shifts

YAO Yizhou¹,
DENG Wen^{2
, ,},
LI Baoguo³

1.
Gaoling Training Base of the People's Armed Police, Beijing 101507, China
2.
People's Liberation Army Unit 63892, Luoyang 471003, China
3.
College of Electronic Science, National University of Defense Technology, Changsha 410073, China

摘要

摘要: 现有基于监督学习范式的短波信号调制识别方法通常假设训练数据(源域)与测试数据(目标域)服从同一分布。然而，短波信道易受电离层变化影响，导致域间分布差异显著，进而引发模型性能退化。此外，无人平台边缘侧部署还面临设备资源受限、标注样本稀缺以及数据隐私保护等多重挑战。针对上述问题，本文提出一种基于源模型迁移的轻量化识别方法，能够在不访问源域数据的条件下实现隐私安全的模型迁移。该方法的优势主要体现在三个方面：首先，具备良好的轻量化特性，仅需在无标签目标域上进行单轮训练即可快速收敛，显著降低了计算开销；其次，具备优异的小样本适应能力，在目标域样本极少的场景下仍能保持较高的识别精度；最后，通过融合涵盖同相/正交分量、幅度/相位信息及频谱特征的多模态信号特征，充分利用特征互补性增强了模型鲁棒性。仿真实验结果表明，该方法在大幅降低资源消耗的同时，在小样本条件下仍能保持稳定的识别性能，验证了其兼具快速收敛、低资源需求和良好泛化能力的特性。
- 短波通信 /
- 调制识别 /
- 源模型迁移 /
- 轻量化 /
- 多模态融合 /
- 小样本
Abstract: Objective Existing short-wave signal modulation recognition methods based on the supervised learning paradigm typically assume that training data (source domain) and test data (target domain) follow identical distributions. However, short-wave channels are susceptible to ionospheric variations, leading to significant distribution discrepancies across domains, which consequently causes model performance degradation. Furthermore, deployment on the edge side of unmanned platforms is constrained by limited device resources, scarce labeled samples, and data privacy requirements. To address these challenges, a lightweight recognition method based on source-model transfer is proposed in this paper, enabling privacy-preserving model adaptation without the need to access source domain data. Methods A multi-modal source-model transfer framework (M-SMOT) is developed, which utilizes information maximization loss and self-supervised pseudo-labeling techniques to facilitate model adaptation without revisiting source domain data. This approach achieves effective cross-channel recognition of short-wave modulation signals while reducing computational resource consumption and preserving data privacy. Additionally, multi-modal information—comprising in-phase/quadrature (I/Q) components, amplitude-phase (AP) characteristics, and spectral features—is fused to leverage complementary feature representations, thereby enhancing the robustness of the recognition network against complex channel variations. Results and Discussions Experimental results demonstrate that the recognition performance of the proposed method consistently surpasses that of the Source-Only baseline across six cross-channel scenarios, with improvements ranging from 0.31% to 10.81% (Table 1). In terms of few-shot adaptation, average recognition accuracies are maintained at 98.3% and 96% relative to the full-sample baseline, even when target domain training samples are reduced to 10% and 1%, respectively (Fig. 12). Ablation studies verify the necessity and effectiveness of the self-supervised pseudo-labeling module (Fig. 16) and the multi-modal fusion strategy (Fig. 17), confirming that both components contribute to the overall performance. Furthermore, the lightweight advantages are quantified: the method requires zero storage for source data, exhibits a peak memory consumption of only 6.00 MB, and achieves convergence within a single fine-tuning epoch (Table 2). These findings validate the capability of the proposed mechanism to mitigate domain discrepancies and protect privacy under resource-constrained conditions. Conclusions The M-SMOT method successfully integrates data privacy protection, source model adaptation, few-shot generalization, and low resource consumption. Consequently, it provides a practical solution for cross-channel modulation recognition in short-wave communications, demonstrating significant potential for deployment on resource-limited edge devices.
- Short-wave communication /
- Modulation recognition /
- Source-model transfer /
- Lightweight /
- Multi-modal fusion /
- Few-shot

HTML全文

图 1 同相正交分量

下载: 全尺寸图片幻灯片

图 2 频谱幅度

下载: 全尺寸图片幻灯片

图 3 二次方谱

下载: 全尺寸图片幻灯片

图 4 幅度相位

下载: 全尺寸图片幻灯片

图 5 多模态源模型迁移框架处理流程示意

下载: 全尺寸图片幻灯片

图 6 低纬度静态环境信道向中等环境信道迁移识别率

下载: 全尺寸图片幻灯片

图 7 低纬度静态环境信道向恶劣环境信道迁移识别率

下载: 全尺寸图片幻灯片

图 8 中纬度静态环境信道向中等环境信道迁移识别率

下载: 全尺寸图片幻灯片

图 9 中纬度静态环境信道向恶劣环境信道迁移识别率

下载: 全尺寸图片幻灯片

图 10 高纬度静态环境信道向中等环境信道迁移识别率

下载: 全尺寸图片幻灯片

图 11 高纬度静态环境信道向恶劣环境信道迁移识别率

下载: 全尺寸图片幻灯片

图 12 低纬度静态环境信道向中等环境信道迁移情况下，不同训练样本量下M-SMOT的识别准确率变化曲线

下载: 全尺寸图片幻灯片

图 13 M-SMOT目标域100%训练样本量平均识别准确率混淆矩阵

下载: 全尺寸图片幻灯片

图 14 M-SMOT目标域10%训练样本量平均识别准确率混淆矩阵

下载: 全尺寸图片幻灯片

图 15 M-SMOT目标域1%训练样本量平均识别准确率混淆矩阵

下载: 全尺寸图片幻灯片

图 16 低纬度静态环境信道向中等环境信道迁移情况下，具有和不具有自监督伪标签的不同算法识别率对比

下载: 全尺寸图片幻灯片

图 17 低纬度静态环境信道向中等环境信道迁移情况下，输入不同模态特征组合的识别率对比

下载: 全尺寸图片幻灯片

表 1 不同跨信道场景下各方法的平均识别准确率(%)

纬度	源域信道状态	目标域信道状态	有监督学习	Source-Only	M-SMOT
低纬度	静态环境	中等环境	98.0	82.7	93.6
低纬度	静态环境	恶劣环境	98.7	96.5	96.8
中纬度	静态环境	中等环境	98.7	88.2	92.9
中纬度	静态环境	恶劣环境	98.9	83.4	94.0
高纬度	静态环境	中等环境	98.7	91.9	97.9
高纬度	静态环境	恶劣环境	98.8	94.8	96.8

下载: 导出CSV

表 2 不同模型开销对比分析

模型	是否需要存储源域数据	单样本推理峰值内存(MB)	目标域数据训练稳定轮次
ResNeXt-DANN^[12]	是	5.1	50
域适应神经网络识别模型^[13]	是	168.3	30
ATLA^[14]	是	10.8	30
AMRMIDAN^[15]	是	5.2	50
M-SMOT(所提模型)	否	6.0	1

下载: 导出CSV

参考文献(18)

[1]	郑庆河, 李秉霖, 于治国, 等. 深度学习使能的自动调制分类技术研究进展[J]. 电子与信息学报, 2025, 47(11): 4096–4111. doi: 10.11999/JEIT250674. ZHENG Qinghe, LI Binglin, YU Zhiguo, et al. Research progress of deep learning enabled automatic modulation classification technology[J]. Journal of Electronics & Information Technology, 2025, 47(11): 4096–4111. doi: 10.11999/JEIT250674.
[2]	O’SHEA T J, CORGAN J, and CLANCY T C. Convolutional radio modulation recognition networks[C]. Proceedings of the 17th International Conference on Engineering Applications of Neural Networks, Aberdeen, UK, 2016: 213–226. doi: 10.1007/978-3-319-44188-7_16.
[3]	O’SHEA T J, ROY T, and CLANCY T C. Over-the-air deep learning based radio signal classification[J]. IEEE Journal of Selected Topics in Signal Processing, 2018, 12(1): 168–179. doi: 10.1109/JSTSP.2018.2797022.
[4]	SUN Xiaochuan, WU Changcheng, LI Yingqi, et al. Toward interference-tolerant automatic modulation recognition via multi-stage feature extraction network[J]. IEEE Transactions on Vehicular Technology, 2025, 74(8): 12629–12640. doi: 10.1109/TVT.2025.3555769.
[5]	郑庆河, 刘方霖, 余礼苏, 等. 基于改进Kolmogorov-Arnold混合卷积神经网络的调制识别方法[J]. 电子与信息学报, 2025, 47(8): 2584–2597. doi: 10.11999/JEIT250161. ZHENG Qinghe, LIU Fanglin, YU Lisu, et al. An improved modulation recognition method based on hybrid Kolmogorov-Arnold convolutional neural network[J]. Journal of Electronics & Information Technology, 2025, 47(8): 2584–2597. doi: 10.11999/JEIT250161.
[6]	LAI Fan, WANG Chengxiang, HUANG Jie, et al. A novel 3D non-stationary massive MIMO channel model for shortwave communication systems[J]. IEEE Transactions on Communications, 2023, 71(9): 5473–5486. doi: 10.1109/TCOMM.2023.3281875.
[7]	李国军, 马欢, 叶昌荣, 等. 基于Watterson模型的短波航空移动信道建模与仿真[J]. 电子与信息学报, 2022, 44(1): 332–338. doi: 10.11999/JEIT200890. LI Guojun, MA Huan, YE Changrong, et al. Modeling and simulation of HF aeronautical mobile channel based on Watterson model[J]. Journal of Electronics & Information Technology, 2022, 44(1): 332–338. doi: 10.11999/JEIT200890.
[8]	ZHUANG Fuzhen, QI Zhiyuan, DUAN Keyu, et al. A comprehensive survey on transfer learning[J]. Proceedings of the IEEE, 2021, 109(1): 43–76. doi: 10.1109/JPROC.2020.3004555.
[9]	LONG Mingsheng, CAO Yue, WANG Jianmin, et al. Learning transferable features with deep adaptation networks[C]. Proceedings of the 32nd International Conference on Machine Learning, Lille, France, 2015: 97–105.
[10]	TZENG E, HOFFMAN J, SAENKO K, et al. Adversarial discriminative domain adaptation[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, 2017: 2962–2971. doi: 10.1109/CVPR.2017.316.
[11]	ZHOU Quan, ZHANG Ronghui, MU Junsheng, et al. AMCRN: Few-shot learning for automatic modulation classification[J]. IEEE Communications Letters, 2022, 26(3): 542–546. doi: 10.1109/LCOMM.2021.3135688.
[12]	许华, 苟泽中, 蒋磊, 等. 适用于样本分布差异的迁移学习调制识别算法[J]. 华中科技大学学报: 自然科学版, 2021, 49(4): 127–132. doi: 10.13245/j.hust.210422. XU Hua, GOU Zezhong, JIANG Lei, et al. Transfer learning modulation recognition algorithm for differences in sample distribution[J]. Journal of Huazhong University of Science & Technology: Natural Science Edition, 2021, 49(4): 127–132. doi: 10.13245/j.hust.210422.
[13]	史蕴豪, 许华, 单俊杰. 基于域适应神经网络的调制方式分类方法[J]. 空军工程大学学报: 自然科学版, 2020, 21(5): 69–75. doi: 10.3969/j.issn.1009-3516.2020.05.011. SHI Yunhao, XU Hua, and SHAN Junjie. A modulation recognition method based on domain adaptive neural network[J]. Journal of Air Force Engineering University: Natural Science Edition, 2020, 21(5): 69–75. doi: 10.3969/j.issn.1009-3516.2020.05.011.
[14]	BU Ke, HE Yuan, JING Xiaojun, et al. Adversarial transfer learning for deep learning based automatic modulation classification[J]. IEEE Signal Processing Letters, 2020, 27: 880–884. doi: 10.1109/LSP.2020.2991875.
[15]	DENG Wen, WANG Xiang, HUANG Zhitao, et al. Modulation classifier: A few-shot learning semi-supervised method based on multimodal information and domain adversarial network[J]. IEEE Communications Letters, 2023, 27(2): 576–580. doi: 10.1109/LCOMM.2022.3225566.
[16]	王祯, 刘伟, 卢万杰, 等. 面向低信噪比序列的多模态联合自动调制方式识别方法[J]. 电子与信息学报, 2025, 47(12): 5082–5093. doi: 10.11999/JEIT250594. WANG Zhen, LIU Wei, LU Wanjie, et al. Multi-modal joint automatic modulation recognition method towards low SNR sequences[J]. Journal of Electronics & Information Technology, 2025, 47(12): 5082–5093. doi: 10.11999/JEIT250594.
[17]	王华华, 张睿哲, 黄永洪. 基于生成式对抗网络和多模态注意力机制的扩频与常规调制信号识别方法[J]. 电子与信息学报, 2024, 46(4): 1212–1221. doi: 10.11999/JEIT230518. WANG Huahua, ZHANG Ruizhe, and HUANG Yonghong. Spread spectrum and conventional modulation signal recognition method based on generative adversarial network and multi-modal attention mechanism[J]. Journal of Electronics & Information Technology, 2024, 46(4): 1212–1221. doi: 10.11999/JEIT230518.
[18]	WU Xueyu, TYRRELL A M, and KO Y. Federated K-means clustering for adaptive OFDM-IM[J]. IEEE Communications Letters, 2023, 27(10): 2648–2651. doi: 10.1109/LCOMM.2023.3308334.