| 73 | 0 | 43 |
| 下载次数 | 被引频次 | 阅读次数 |
该文提出了一种基于域适应的辐射源个体识别方法，旨在解决传统方法在不同域样本下识别准确率低的问题。该方法引入一对多(One-Vs-All, OVA)分类器，为每个类别训练一个二分类器，通过最小化正类与其最近负类之间的熵值，压缩正类空间以增强预测可靠性。同时，结合闭集分类器和OVA分类器的输出，设计了一致性评价和样本均衡策略，用于筛选高质量伪标签，优化目标域样本的模型调整。在Oracle和Wisig数据集上进行的实验表明，该方法在多种噪声条件下均优于现有方法。在-3 dB噪声条件下，Oracle数据集的准确率达到69.61%,Wisig数据集的准确率达到73.55%,相比其他方法分别提升了约1～8百分点和5～18百分点；在多种噪声条件下，Oracle数据集平均准确率达到87.36%,Wisig数据集平均准确率达到89.11%,相比对比方法高出约1～3百分点。该方法在高噪声环境下展现出更强的鲁棒性和适应性，为辐射源个体识别提供了有效的解决方案。
Abstract：We propose a domain adaptation-based method for radiation source individual recognition, aiming to address the issue of low recognition accuracy of traditional methods under different domain samples. The method introduces a One-Vs-All(OVA) classifier, training a binary classifier for each category. By minimizing the entropy between the positive class and its nearest negative class, the positive class space is compressed to enhance prediction reliability. In addition, a consistency evaluation and sample balancing strategy is designed by combining the outputs of the closed-set classifier and the OVA classifier. These strategies are used to filter high-quality pseudo-labels and optimize the model adaptation for target domain samples. Experiments are conducted on the Oracle and Wisig datasets, which show that the proposed method outperforms existing methods under various noise conditions. Under-3 dB noise conditions, the accuracy on the Oracle dataset reaches 69.61%,and on the Wisig dataset, it reaches 73.55%,representing improvements of approximately 1～8 percentage points and 5～18 percentage points compared to other methods, respectively. Under multiple noise conditions, the average accuracy on the Oracle dataset is 87.36%,and on the Wisig dataset, it is 89.11%,which are about 1～3 percentage points higher than that of the comparison methods. The proposed method demonstrates stronger robustness and adaptability in high-noise environments, providing an effective solution for radiation source individual recognition.
[1]SCHERBATYUK A P.The AUV positioning using ranges from one transponder LBL[C]//Challenges of our changing global environment.San Diego:IEEE,1995:1620-1623.
[2]WAWRZY N'SKI W,ZIEJA M,ZOKOWSKI M,et al.Optimization of autonomous underwater vehicle mission planning process[J].Bulletin of the Polish Academy of Sciences Technical Sciences,2022,70(2):140371.
[3]URETEN O,SERINKEN N.Wireless security through RFfingerprinting Securit e sans fil par une empreinte digitale RF[J].Journal of Electrical and Computer Engineering,2007,32:27-33.
[4]HUANG H,GUO S,GUI G,et al.Deep learning for physical-layer5G wireless techniques:opportunities,challenges and solutions[J].IEEE Wireless Communications,2020,27(1):214-222.
[5]QU S,ZOU T,HE L,et al.LEAD:learning decomposition for source-free universal domain adaptation[C]//2024IEEE/CVF conference on computer vision and pattern recognition(CVPR).Seattle:IEEE,2024:23334-23343.
[6]ZHANG H,SU Y,XU X,et al.Improving the generalization of segmentation foundation model under distribution shift via weakly supervised adaptation[C]//2024IEEE/CVF conference on computer vision and pattern recognition(CVPR).Seattle:IEEE,2024:23385-23395.
[7]WU Y,HE K,YAN Y,et al.Revisit overconfidence for OOD detection:reassigned contrastive learning with adaptive class-dependent threshold[C]//Proceedings of the2022conference of the North American chapter of the association for computational linguistics:human language technologies.Seattle:Association for Computational Linguistics,2022:4165-4179.
[8]PAN S J,YANG Q.A survey on transfer learning[J].IEEETransactions on Knowledge and Data Engineering,2010,22(10):1345-1359.
[9]TZENG E,HOFFMAN J,ZHANG N,et al.Deep domain confusion:maximizing for domain invariance[J].ar Xiv:1412.3474,2014.
[10]ZHANG Y,LIU T,LONG M,et al.Bridging theory and algorithm for domain adaptation[C]//Proceedings of the36th international conference on machine learning.Long Beach:PMLR,2019:7404-7413.
[11]LONG M,ZHU H,WANG J,et al.Deep transfer learning with joint adaptation networks[C]//Proceedings of the34th international conference on machine learning.[s.l.]:[s.n.],2017:2208-2217.
[12]XIAO Z,WANG H,JIN Y,et al.SPA:a graph spectral alignment perspective for domain adaptation[C]//Advances in neural information processing systems.New Orleans:PMLR,2023:37252-37272.
[13]GOODFELLOW I,POUGET-ABADIE J,MIRZA M,et al.Generative adversarial nets[C/OL]//Advances in neural information processing systems.Montreal:Curran Associates,Inc.,2014:2672-2680.
[14]GANIN Y,LEMPITSKY V.Unsupervised domain adaptation by backpropagation[C]//Proceedings of the32nd international conference on machine learning.Lille:PMLR,2015:1180-1189.
[15]张涛涛,谢钧,乔平娟.基于多源无监督域适应的辐射源个体识别方法[J].计算机与现代化,2025(3):45-51.
[16]GHIFARY M,KLEIJN W B,ZHANG M,et al.Deep reconstruction-classification networks for unsupervised domain adaptation[C]//Computer vision-ECCV2016.Amsterdam:Springer,2016:597-613.
[17]SANKARANARAYANAN S,BALAJI Y,CASTILLO C D,et al.Generate to adapt:aligning domains using generative adversarial networks[C]//2018IEEE/CVF conference on computer vision and pattern recognition.Salt Lake City:IEEE,2018:8503-8512.
[18]WOLF D A,GALIN R R.Performance comparative analysis of Ov A,Av A,and Ov O algorithms in multi-class classification[C]//2024 17th international conference on management of large-scale system development(MLSD).Moscow:IEEE,2024:1-5.
[19]SANKHE K,BELGIOVINE M,ZHOU F,et al.ORACLE:optimized radio classification through convolutional neural networks[C]//IEEE INFOCOM2019-IEEE conference on computer communications.Paris:IEEE,2019:370-378.
[20]HANNA S,KARUNARATNE S,CABRIC D.Wi Sig:a large-scale wifi signal dataset for receiver and channel agnostic RF fingerprinting[J].IEEE Access,2022,10:22808-22818.
基本信息:
DOI：10.20165/j.cnki.ISSN1673-629X.2025.0219
中图分类号:TP18
引用信息:
[1]王闯,俞璐,潘成,等.基于OVA与伪标签优化的辐射源域适应识别方法[J].计算机技术与发展,2026,36(01):73-79.DOI:10.20165/j.cnki.ISSN1673-629X.2025.0219.
基金信息:
国家自然基金项目(62471486)
2025-08-14
2025-08-14
2025-08-14