Article(id=1251856525644481271, tenantId=1146029695717560320, journalId=1251234268282663017, issueId=1251856520619700745, articleNumber=null, orderNo=null, doi=10.3969/j.issn.1003-3106.2025.11.003, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1751817600000, receivedDateStr=2025-07-07, revisedDate=null, revisedDateStr=null, acceptedDate=null, acceptedDateStr=null, onlineDate=1776395454083, onlineDateStr=2026-04-17, pubDate=1762272000000, pubDateStr=2025-11-05, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1776395454083, onlineIssueDateStr=2026-04-17, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1776395454083, creator=13701087609, updateTime=1776395454083, updator=13701087609, issue=Issue{id=1251856520619700745, tenantId=1146029695717560320, journalId=1251234268282663017, year='2025', volume='55', issue='11', pageStart='2131', pageEnd='2324', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=1, specialIssue=null, createTime=1776395452885, creator=13701087609, updateTime=1776395571911, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1251857019939013255, tenantId=1146029695717560320, journalId=1251234268282663017, issueId=1251856520619700745, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1251857019939013256, tenantId=1146029695717560320, journalId=1251234268282663017, issueId=1251856520619700745, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=2153, endPage=2162, ext={EN=ArticleExt(id=1251856527393506055, articleId=1251856525644481271, tenantId=1146029695717560320, journalId=1251234268282663017, language=EN, title=Modulation Recognition Algorithm Based on Unsupervised Domain Adaptation, columnId=1251856523492798993, journalTitle=Radio Engineering, columnName=Signal and Information Processing, runingTitle=null, highlight=null, articleAbstract=

Modulation recognition is a critical task in wireless communications. Although deep learning methods have achieved remarkable progress in this field, they still face the challenge of insufficient generalization ability in complex non-cooperative environments—particularly when confronted with varying channel conditions, which can obscure the subtle discriminative features between structurally similar modulation schemes (e. g. 16QAM and 64QAM) and thus degrade recognition performance. To address this unique challenge in the field of modulation recognition, an unsupervised adversarial domain adaptation method named Feature Alignment and Discrimination Domain Adaptation ( FADDA) is proposed. The core of FADDA is the introduction of a contrastive learning-based feature alignment loss on the basis of adversarial training. Adversarial training is responsible for learning domain-invariant features to adapt to channel variations, while the feature alignment. loss fundamentally enhances the model' s ability to distinguish between easily confused modulation types by explicitly reinforcing the compactness of intra-class features and the separability of inter-class features. Experimental results show that without target-domain labels, this method can significantly improve the model's cross-channel modulation recognition performance and demonstrate strong generalization ability.

, correspAuthors=Meiying WEI, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Shuo CHANG, Shun XU, Meiying WEI), CN=ArticleExt(id=1251856538646823131, articleId=1251856525644481271, tenantId=1146029695717560320, journalId=1251234268282663017, language=CN, title=基于无监督领域自适应的调制识别算法, columnId=1251856523803177493, journalTitle=无线电工程, columnName=信号与信息处理, runingTitle=null, highlight=null, articleAbstract=

调制识别是无线通信中的关键任务,深度学习方法虽已取得显著进展,但在复杂非合作环境下仍面临泛化能力不足的挑战。特别是,变化的信道条件会模糊结构相似的调制方式(如16QAM与64QAM)之间细微的判别性特征,导致识别性能下降。针对这一调制识别领域的特有挑战,提出特征对齐对抗领域自适应(Feature Alignment and Discrimination Domain Adaptation,FADDA)的无监督对抗领域自适应算法。该方法的核心是在对抗训练的基础上,创新性地引入了基于对比学习的特征对齐损失。对抗训练负责学习域不变特征以适应信道变化,特征对齐损失则通过显式地增强类内特征的紧凑性和类间特征的可分性,从而在根本上提升模型对易混淆调制类型的分辨能力。实验验证表明,该方法在缺乏目标域标签的情况下,能够显著增强模型的跨信道调制识别性能,展现出强大的泛化能力。

, correspAuthors=魏梅英, authorNote=null, correspAuthorsNote=
魏梅英 女,(1985—),硕士,高级工程师。主要研究方向:无线电监测、无线电干扰定位、信号分析。
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=QB60+9jV1KjLxNW+Oj48Sg==, magXml=1qVZNubpF1jfMEETHFgFkQ==, pdfUrl=null, pdf=7jgc0Wmf0+/FsIKuwFbiBw==, pdfFileSize=8122829, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=ACzXppQd+KLE2Ham+tmcYw==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=GshMgCtzyEiIoe0lq4LRBQ==, mapNumber=null, authorCompany=null, fund=null, authors=

昌硕 男,(1993—),博士,讲师,硕士生导师。主要研究方向:频谱深度认知,多源信息融合,深度学习。

胥顺 男,(2000—),硕士研究生。主要研究方向:智能信号处理、自动调制识别、深度学习。

, authorsList=昌硕, 胥顺, 魏梅英)}, authors=[Author(id=1251856540790112507, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1251856540936913157, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, authorId=1251856540790112507, language=EN, stringName=Shuo CHANG, firstName=Shuo, middleName=null, lastName=CHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1.School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing 100876, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1251856541079519502, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, authorId=1251856540790112507, language=CN, stringName=昌硕, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1.北京邮电大学 网络空间安全学院,北京 100876, bio={"content":"

昌硕 男,(1993—),博士,讲师,硕士生导师。主要研究方向:频谱深度认知,多源信息融合,深度学习。

"}, bioImg=null, bioContent=

昌硕 男,(1993—),博士,讲师,硕士生导师。主要研究方向:频谱深度认知,多源信息融合,深度学习。

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1251856539024310504, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, xref=1., ext=[AuthorCompanyExt(id=1251856539032699113, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856539024310504, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1.School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing 100876, China), AuthorCompanyExt(id=1251856539036893418, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856539024310504, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1.北京邮电大学 网络空间安全学院,北京 100876)])]), Author(id=1251856541163405585, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1251856541297623325, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, authorId=1251856541163405585, language=EN, stringName=Shun XU, firstName=Shun, middleName=null, lastName=XU, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1.School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing 100876, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1251856541381509410, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, authorId=1251856541163405585, language=CN, stringName=胥顺, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1.北京邮电大学 网络空间安全学院,北京 100876, bio={"content":"

胥顺 男,(2000—),硕士研究生。主要研究方向:智能信号处理、自动调制识别、深度学习。

"}, bioImg=null, bioContent=

胥顺 男,(2000—),硕士研究生。主要研究方向:智能信号处理、自动调制识别、深度学习。

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1251856539024310504, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, xref=1., ext=[AuthorCompanyExt(id=1251856539032699113, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856539024310504, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1.School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing 100876, China), AuthorCompanyExt(id=1251856539036893418, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856539024310504, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1.北京邮电大学 网络空间安全学院,北京 100876)])]), Author(id=1251856541477978409, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1251856541570253103, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, authorId=1251856541477978409, language=EN, stringName=Meiying WEI, firstName=Meiying, middleName=null, lastName=WEI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, address=2.State Radio Regulation of China, Beijing 100086, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1251856541666722101, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, authorId=1251856541477978409, language=CN, stringName=魏梅英, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, address=2.国家无线电监测中心,北京 100086, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1251856540639117553, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, xref=2., ext=[AuthorCompanyExt(id=1251856540655894771, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856540639117553, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2.State Radio Regulation of China, Beijing 100086, China), AuthorCompanyExt(id=1251856540672671988, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856540639117553, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2.国家无线电监测中心,北京 100086)])])], keywords=[Keyword(id=1251856541809328446, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, orderNo=1, keyword=modulation recognition), Keyword(id=1251856541884825922, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, orderNo=2, keyword=unsupervised learning), Keyword(id=1251856541985489225, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, orderNo=3, keyword=domain adaptation), Keyword(id=1251856542119706958, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, orderNo=1, keyword=调制识别), Keyword(id=1251856542224564569, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, orderNo=2, keyword=无监督学习), Keyword(id=1251856542333616478, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, orderNo=3, keyword=域自适应)], refs=[Reference(id=1251856546280456682, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2019, volume=34, issue=3, pageStart=134, pageEnd=142, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=SAAD W, BENNIS M, CHEN M Z, journalName=IEEE Network, refType=null, unstructuredReference=SAAD W, BENNIS M, CHEN M Z. A Vision of 6G Wireless Systems: Applications, Trends, Technologies, and Open Research Problems[J]. IEEE Network,2019, 34(3):134-142., articleTitle=A Vision of 6G Wireless Systems: Applications, Trends, Technologies, and Open Research Problems, refAbstract=null), Reference(id=1251856546452423151, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2019, volume=57, issue=8, pageStart=84, pageEnd=90, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=LETAIEF K B, CHEN W, SHI Y M, journalName=IEEE Communications Magazine, refType=null, unstructuredReference=LETAIEF K B, CHEN W, SHI Y M, et al. The Roadmap to 6G: AI Empowered Wireless Networks[J]. IEEE Communications Magazine,2019,57(8):84-90., articleTitle=The Roadmap to 6G: AI Empowered Wireless Networks, refAbstract=null), Reference(id=1251856546557280754, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2022, volume=129, issue=null, pageStart=103650, pageEnd=null, url=null, language=null, rfNumber=[3], rfOrder=2, authorNames=ZHANG F X, LUO C B, XU J L, journalName=Digital Signal Processing, refType=null, unstructuredReference=ZHANG F X, LUO C B, XU J L, et al. Deep Learning Based Automatic Modulation Recognition: Models, Datasets, and Challenges[J]. Digital Signal Processing,2022, 129:103650., articleTitle=Deep Learning Based Automatic Modulation Recognition: Models, Datasets, and Challenges, refAbstract=null), Reference(id=1251856546636972536, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2020, volume=69, issue=3, pageStart=3491, pageEnd=3495, url=null, language=null, rfNumber=[4], rfOrder=3, authorNames=WANG Y, YANG J, LIU M, journalName=IEEE Transactions on Vehicular Technology, refType=null, unstructuredReference=WANG Y,YANG J,LIU M,et al. LightAMC: Lightweight Automatic Modulation Classification via Deep Learning and Compressive Sensing[J]. IEEE Transactions on Vehicular Technology,2020,69(3):3491-3495., articleTitle=LightAMC: Lightweight Automatic Modulation Classification via Deep Learning and Compressive Sensing, refAbstract=null), Reference(id=1251856546729247227, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=1995, volume=null, issue=null, pageStart=886, pageEnd=890, url=null, language=null, rfNumber=[5], rfOrder=4, authorNames=HO K C, PROKOPIW W, CHAN Y T, journalName=null, refType=null, unstructuredReference=HO K C, PROKOPIW W, CHAN Y T. Modulation Identification by the Wavelet Transform[C]//Proceedings of MILCOM'95. San Diego:IEEE,1995:886-890., articleTitle=Modulation Identification by the Wavelet Transform, refAbstract=null), Reference(id=1251856546804744703, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2022, volume=41, issue=7, pageStart=126, pageEnd=132, url=null, language=null, rfNumber=[6], rfOrder=5, authorNames=张文娜, 孙运强, 姚爱琴, journalName=国外电子测量技术, refType=null, unstructuredReference=张文娜,孙运强,姚爱琴,.基于高阶累积量的数字调制识别[J].国外电子测量技术,2022,41(7):126-132., articleTitle=基于高阶累积量的数字调制识别, refAbstract=null), Reference(id=1251856546876047874, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2016, volume=38, issue=3, pageStart=674, pageEnd=680, url=null, language=null, rfNumber=[7], rfOrder=6, authorNames=赵雄文, 郭春霞, 李景春, journalName=电子与信息学报, refType=null, unstructuredReference=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[J].电子与信息学报, 2016,38(3):674-680., articleTitle=基于高阶累积量和循环谱的信号调制方式混合识别算法, refAbstract=null), Reference(id=1251856546980905479, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2015, volume=31, issue=8, pageStart=975, pageEnd=985, url=null, language=null, rfNumber=[8], rfOrder=7, authorNames=孔豫京, 黄焱, 马金全, journalName=信号处理, refType=null, unstructuredReference=孔豫京,黄焱,马金全. Alpha稳定分布噪声背景下衰落信号的调制识别方法[J].信号处理,2015,31(8):975-985., articleTitle=Alpha稳定分布噪声背景下衰落信号的调制识别方法, refAbstract=null), Reference(id=1251856547073180170, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2020, volume=8, issue=null, pageStart=67366, pageEnd=67376, url=null, language=null, rfNumber=[9], rfOrder=8, authorNames=ZHOU R L, LIU F G, GRAVELLE C W, journalName=IEEE Access, refType=null, unstructuredReference=ZHOU R L, LIU F G, GRAVELLE C W. Deep Learning for Modulation Recognition: A Survey with a Demonstration[J]. IEEE Access,2020,8:67366-67376., articleTitle=Deep Learning for Modulation Recognition: A Survey with a Demonstration, refAbstract=null), Reference(id=1251856547178037773, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=1, pageEnd=6, url=null, language=null, rfNumber=[10], rfOrder=9, authorNames=KUMAR Y, JAJOO G, YADAV S K, journalName=null, refType=null, unstructuredReference=KUMAR Y, JAJOO G, YADAV S K. 2D-FFT Based Modulation Classification Using Deep Convolution Neural Network[C]//2020 IEEE 17th India Council International Conference (INDICON). New Delhi:IEEE,2020:1-6., articleTitle=2D-FFT Based Modulation Classification Using Deep Convolution Neural Network, refAbstract=null), Reference(id=1251856547253535250, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2023, volume=11, issue=8, pageStart=1632, pageEnd=null, url=null, language=null, rfNumber=[11], rfOrder=10, authorNames=ZHANG R, HE C B, JING L Y, journalName=Journal of Marine Science and Engineering, refType=null, unstructuredReference=ZHANG R, HE C B, JING L Y, et al. A Modulation Recognition System for Underwater Acoustic Communication Signals Based on Higher-order Cumulants and Deep Learning[J]. Journal of Marine Science and Engineering,2023,11(8):1632., articleTitle=A Modulation Recognition System for Underwater Acoustic Communication Signals Based on Higher-order Cumulants and Deep Learning, refAbstract=null), Reference(id=1251856547329032728, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2020, volume=50, issue=9, pageStart=727, pageEnd=731, url=null, language=null, rfNumber=[12], rfOrder=11, authorNames=董聪, 张传武, 高勇, journalName=无线电工程, refType=null, unstructuredReference=董聪,张传武,高勇.基于残差神经网络的通信混合信号识别[J].无线电工程,2020,50(9):727-731., articleTitle=基于残差神经网络的通信混合信号识别, refAbstract=null), Reference(id=1251856547404530203, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2023, volume=53, issue=6, pageStart=1328, pageEnd=1341, url=null, language=null, rfNumber=[13], rfOrder=12, authorNames=杨梓辰, 笪清安, 李思照, journalName=无线电工程, refType=null, unstructuredReference=杨梓辰,笪清安,李思照.面向深度学习模型的对抗攻击与防御综述[J].无线电工程,2023,53(6):1328-1341., articleTitle=面向深度学习模型的对抗攻击与防御综述, refAbstract=null), Reference(id=1251856547488416284, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2022, volume=26, issue=11, pageStart=2646, pageEnd=2650, url=null, language=null, rfNumber=[14], rfOrder=13, authorNames=LIANG Z, XIE J, YANG X, journalName=IEEE Communications Letters, refType=null, unstructuredReference=LIANG Z, XIE J, YANG X, et al. Self-training Based Adversarial Domain Adaptation for Radio Signal Recognition[J].IEEE Communications Letters,2022,26(11):2646-2650., articleTitle=Self-training Based Adversarial Domain Adaptation for Radio Signal Recognition, refAbstract=null), Reference(id=1251856547568108063, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2025-05-11, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[15], rfOrder=14, authorNames=ARBEL M, KORBA A, SALIM A, journalName=null, refType=null, unstructuredReference=ARBEL M, KORBA A, SALIM A, et al. Maximum Mean Discrepancy Gradient Flow[EB/OL]. (2019-06-11)[2025-05-11]. https://arxiv.org/abs/1906.04370., articleTitle=Maximum Mean Discrepancy Gradient Flow, refAbstract=null), Reference(id=1251856547651994147, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2016, volume=null, issue=null, pageStart=443, pageEnd=450, url=null, language=null, rfNumber=[16], rfOrder=15, authorNames=SUN B C, SAENKO K, journalName=null, refType=null, unstructuredReference=SUN B C,SAENKO K. Deep CORAL:Correlation Alignment for Deep Domain Adaptation[C]//European Conference on Computer Vision. Amsterdam:Springer,2016:443-450., articleTitle=Deep CORAL:Correlation Alignment for Deep Domain Adaptation, refAbstract=null), Reference(id=1251856547719103014, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2014, volume=null, issue=null, pageStart=2672, pageEnd=2680, url=null, language=null, rfNumber=[17], rfOrder=16, authorNames=GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, journalName=null, refType=null, unstructuredReference=GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative Adversarial Nets[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems. Montreal:ACM,2014:2672-2680., articleTitle=Generative Adversarial Nets, refAbstract=null), Reference(id=1251856547786211882, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, doi=null, pmid=null, pmcid=null, year=2025-05-11, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[18], rfOrder=17, authorNames=LONG M S, CAO Z J, WANG J M, journalName=null, refType=null, unstructuredReference=LONG M S, CAO Z J, WANG J M, et al. Conditional Adversarial Domain Adaptation[EB/OL]. (2017-05-26)[2025-05-11]. https://arxiv.org/abs/1705.10667., articleTitle=Conditional Adversarial Domain Adaptation, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1251856539024310504, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, xref=1., ext=[AuthorCompanyExt(id=1251856539032699113, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856539024310504, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1.School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing 100876, China), AuthorCompanyExt(id=1251856539036893418, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856539024310504, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1.北京邮电大学 网络空间安全学院,北京 100876)]), AuthorCompany(id=1251856540639117553, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, xref=2., ext=[AuthorCompanyExt(id=1251856540655894771, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856540639117553, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2.State Radio Regulation of China, Beijing 100086, China), AuthorCompanyExt(id=1251856540672671988, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, companyId=1251856540639117553, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2.国家无线电监测中心,北京 100086)])], figs=[ArticleFig(id=1251856542472028518, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Fig.1, caption=Overall structure of FADDA, figureFileSmall=SYIrMzsWuBNl1ais9kOEnQ==, figureFileBig=ACzXppQd+KLE2Ham+tmcYw==, tableContent=null), ArticleFig(id=1251856542551720299, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=图1, caption=FADDA算法整体架构, figureFileSmall=SYIrMzsWuBNl1ais9kOEnQ==, figureFileBig=ACzXppQd+KLE2Ham+tmcYw==, tableContent=null), ArticleFig(id=1251856542765629816, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Fig.2, caption=Featrue extractor network architecture, figureFileSmall=HHxKog0XyX/lnCh2v32D3w==, figureFileBig=FsrAuJuoT1nFC89YW2tanA==, tableContent=null), ArticleFig(id=1251856542878876033, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=图2, caption=特征提取器网络架构, figureFileSmall=HHxKog0XyX/lnCh2v32D3w==, figureFileBig=FsrAuJuoT1nFC89YW2tanA==, tableContent=null), ArticleFig(id=1251856542958567813, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Fig.3, caption=Classifier network architecture, figureFileSmall=Uhbhc+U+FVSVskoWE0Y9Tw==, figureFileBig=VQmXnxEldQgSoIfR8g5Wxg==, tableContent=null), ArticleFig(id=1251856543059231113, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=图3, caption=分类器网络架构, figureFileSmall=Uhbhc+U+FVSVskoWE0Y9Tw==, figureFileBig=VQmXnxEldQgSoIfR8g5Wxg==, tableContent=null), ArticleFig(id=1251856543143117200, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Fig.4, caption=Discriminator network architecture, figureFileSmall=DuZ4t6X63ehP2l6PAhQX7Q==, figureFileBig=LIUkxrYXKgTkjc+KFQzl6A==, tableContent=null), ArticleFig(id=1251856543214420373, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=图4, caption=判别器网络架构, figureFileSmall=DuZ4t6X63ehP2l6PAhQX7Q==, figureFileBig=LIUkxrYXKgTkjc+KFQzl6A==, tableContent=null), ArticleFig(id=1251856543327666588, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Fig.5, caption=Classification accuracy of domain adaptation algorithms under different SNRs, figureFileSmall=bkIfaBxdkRc/XWGDGYYHLA==, figureFileBig=uQ1WTPRGrkWE1wPQw0oIGQ==, tableContent=null), ArticleFig(id=1251856543407358369, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=图5, caption=领域自适应算法不同SNR下分类精度, figureFileSmall=bkIfaBxdkRc/XWGDGYYHLA==, figureFileBig=uQ1WTPRGrkWE1wPQw0oIGQ==, tableContent=null), ArticleFig(id=1251856544996999591, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Fig.6, caption=Comparison of confusion matrices before and after domain adaptation at 0 dB, figureFileSmall=eERmP8VF/XbRZzpskCMd8Q==, figureFileBig=9BE72H+uZE+ie2zWIOGWSQ==, tableContent=null), ArticleFig(id=1251856545085079983, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=图6, caption=0 dB领域自适应前后混淆矩阵对比, figureFileSmall=eERmP8VF/XbRZzpskCMd8Q==, figureFileBig=9BE72H+uZE+ie2zWIOGWSQ==, tableContent=null), ArticleFig(id=1251856545164771764, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Fig.7, caption=Comparison of confusion matrices before and after domain adaptation at 8 dB, figureFileSmall=Ikm0jEGstrlhQt/fh+ZMkQ==, figureFileBig=PajX25YwXP0GkZICd9hD9A==, tableContent=null), ArticleFig(id=1251856545252852154, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=图7, caption=8 dB领域自适应前后混淆矩阵对比, figureFileSmall=Ikm0jEGstrlhQt/fh+ZMkQ==, figureFileBig=PajX25YwXP0GkZICd9hD9A==, tableContent=null), ArticleFig(id=1251856545328349628, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Tab.1, caption=

Parameter configurations of the dataset

, figureFileSmall=null, figureFileBig=null, tableContent=
参数
数据格式[2×1024]的IQ信号
样本总量143000
SNR范围/dB(步长2 dB)[-6,18]
单调制类型单SNR样本数1000
采样率/kHz200
每个样本符号数128
滚降系数0.35
每符号采样点数8
滤波器类型升余弦滚降滤波器
), ArticleFig(id=1251856545416430018, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=表1, caption=

数据集的参数配置

, figureFileSmall=null, figureFileBig=null, tableContent=
参数
数据格式[2×1024]的IQ信号
样本总量143000
SNR范围/dB(步长2 dB)[-6,18]
单调制类型单SNR样本数1000
采样率/kHz200
每个样本符号数128
滚降系数0.35
每符号采样点数8
滤波器类型升余弦滚降滤波器
), ArticleFig(id=1251856545512899019, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Tab.2, caption=

Parameter configurations of Rayleigh channel

, figureFileSmall=null, figureFileBig=null, tableContent=
参数
路径延迟/s[0,0.2,0.6,1.1,1.5]×10-6
路径增益/dB[0,-3,-5,-8,-12]
最大多普勒频移/Hz8
), ArticleFig(id=1251856545630339535, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=表2, caption=

Rayleigh信道参数配置

, figureFileSmall=null, figureFileBig=null, tableContent=
参数
路径延迟/s[0,0.2,0.6,1.1,1.5]×10-6
路径增益/dB[0,-3,-5,-8,-12]
最大多普勒频移/Hz8
), ArticleFig(id=1251856545722614229, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Tab.3, caption=

Parameter configurations of Rician channel

, figureFileSmall=null, figureFileBig=null, tableContent=
参数
K因子12
路径延迟/s[0,0.1,0.25,0.4]×10-6
路径增益/dB[0,-3,-5,-8,-12]
最大多普勒频移/Hz8
), ArticleFig(id=1251856545827471832, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=表3, caption=

Rician信道参数配置

, figureFileSmall=null, figureFileBig=null, tableContent=
参数
K因子12
路径延迟/s[0,0.1,0.25,0.4]×10-6
路径增益/dB[0,-3,-5,-8,-12]
最大多普勒频移/Hz8
), ArticleFig(id=1251856545907163611, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Tab.4, caption=

Performance comparison before and aftermigration under different channels

, figureFileSmall=null, figureFileBig=null, tableContent=
模型AWGN→RayleighAWGN→RicianRayleigh→Rician
上界0.9140.9180.918
DANN0.8450.8270.699
FADDA0.8650.8340.719
下界0.6270.5080.366
), ArticleFig(id=1251856545986855391, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=表4, caption=

不同信道下迁移前后性能对比

, figureFileSmall=null, figureFileBig=null, tableContent=
模型AWGN→RayleighAWGN→RicianRayleigh→Rician
上界0.9140.9180.918
DANN0.8450.8270.699
FADDA0.8650.8340.719
下界0.6270.5080.366
), ArticleFig(id=1251856546070741472, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=EN, label=Tab.5, caption=

Performance comparison before and aftermigration under different channels

, figureFileSmall=null, figureFileBig=null, tableContent=
模型平均分类精度
下界0.627
DANN0.845
Ours (w/o Ld0.821
FADDA0.865
), ArticleFig(id=1251856546158821861, tenantId=1146029695717560320, journalId=1251234268282663017, articleId=1251856525644481271, language=CN, label=表5, caption=

不同信道下迁移前后性能对比

, figureFileSmall=null, figureFileBig=null, tableContent=
模型平均分类精度
下界0.627
DANN0.845
Ours (w/o Ld0.821
FADDA0.865
)], attaches=null, journal=Journal(id=1251231494547484770, delFlag=0, nameCn=无线电工程, nameEn=Radio Engineering, nameHistory1=null, nameHistory2=null, issn=1003-3106, eissn=, cn=13-1097/TN, coden=null, periodic=月刊, language=CN, oaType=1, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=, officePhone=, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=g0Lcfmfo7oUYDLtQTsujLQ==, journalPrice=null, startedYear=null, abbrevIsoEn=Radio Engineering, journalRemark=null, publicationField=null, createdTime=1776246435060, updatedTime=1776252047215, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=R, firstLetterEn=R, subjectCode=Engineering, subjectName=工程, subjectCodeEn=Engineering, subjectNameEn=null, picCn=g0Lcfmfo7oUYDLtQTsujLQ==, picEn=fe/NRE6hx/Z5ZHweFTw2gA==, jcr=null, cjcr=null, exts=[JournalExt(id=1251255033736676338, language=CN, name=无线电工程, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1776252047237, updatedTime=1776252047237, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://wxdg.cbpt.cnki.net/index.aspx?t=1, submissionEditorUrl=https://wxdg.cbpt.cnki.net/index.aspx?t=3, submissionReviewUrl=https://wxdg.cbpt.cnki.net/index.aspx?t=2, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1251255033787007987, language=EN, name=Radio Engineering, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1776252047249, updatedTime=1776252047249, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://wxdg.cbpt.cnki.net/index.aspx?t=1, submissionEditorUrl=https://wxdg.cbpt.cnki.net/index.aspx?t=3, submissionReviewUrl=https://wxdg.cbpt.cnki.net/index.aspx?t=2, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1251234268282663017, websiteList=[Website(id=1251257283494232110, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1251234268282663017, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/wxdgc/CN, language=CN, createTime=1776252583625, createBy=18614031015, updateTime=1776253601640, updateBy=18614031015, name=无线电工程-中文, tplId=1146099689490845704, title=无线电工程, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1251261682236997912, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=articleTextType, value=kx, createTime=1776253632363, updateTime=1776253632363, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682199249173, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=banner, value=null, createTime=1776253632354, updateTime=1776253632354, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682262163739, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=grayFlag, value=0, createTime=1776253632369, updateTime=1776253632369, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682190860564, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=logo, value=https://castjournals.cast.org.cn/joweb/wxdgc/CN/file/pic?fileId=AnvFcVcgnutwkVR+TFSYqg==, createTime=1776253632352, updateTime=1776253632352, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682283135261, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=minRunFlag, value=0, createTime=1776253632374, updateTime=1776253632374, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682224414999, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/wxdgc/CN/file/pic, createTime=1776253632360, updateTime=1776253632360, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682274746652, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=silenceFlag, value=0, createTime=1776253632372, updateTime=1776253632372, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682207637782, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1776253632356, updateTime=1776253632356, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682245386521, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=themeColor, value=null, createTime=1776253632365, updateTime=1776253632365, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261682249580826, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283494232110, code=themeStyle, value=null, createTime=1776253632366, updateTime=1776253632366, creator=18614031015, updator=18614031015)]), Website(id=1251257283603284042, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1251234268282663017, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/wxdgc/EN, language=EN, createTime=1776252583647, createBy=18614031015, updateTime=1776253597767, updateBy=18614031015, name=无线电工程-英文, tplId=1146101810881728533, title=Radio Engineering, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1251261640977625682, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=articleTextType, value=kx, createTime=1776253622526, updateTime=1776253622526, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261640956654159, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=banner, value=null, createTime=1776253622521, updateTime=1776253622521, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261641002791509, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=grayFlag, value=0, createTime=1776253622532, updateTime=1776253622532, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261640948265550, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=logo, value=https://castjournals.cast.org.cn/joweb/wxdgc/EN/file/pic?fileId=AnvFcVcgnutwkVR+TFSYqg==, createTime=1776253622519, updateTime=1776253622519, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261641015374423, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=minRunFlag, value=0, createTime=1776253622535, updateTime=1776253622535, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261640969237073, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/wxdgc/EN/file/pic, createTime=1776253622524, updateTime=1776253622524, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261641011180118, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=silenceFlag, value=0, createTime=1776253622534, updateTime=1776253622534, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261640965042768, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1776253622523, updateTime=1776253622523, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261640986014291, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=themeColor, value=null, createTime=1776253622528, updateTime=1776253622528, creator=18614031015, updator=18614031015), WebsiteProps(id=1251261640994402900, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1251257283603284042, code=themeStyle, value=null, createTime=1776253622530, updateTime=1776253622530, creator=18614031015, updator=18614031015)])], journalTitle=无线电工程, weixinUrl=null, journalUrl=https://wxdg.cbpt.cnki.net/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Radio Engineering, journalPhotoCn=g0Lcfmfo7oUYDLtQTsujLQ==, journalPhotoEn=fe/NRE6hx/Z5ZHweFTw2gA==, journalFirstLetter=R, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/wxdgc/CN/10.3969/j.issn.1003-3106.2025.11.003, detailUrlEn=https://castjournals.cast.org.cn/joweb/wxdgc/EN/10.3969/j.issn.1003-3106.2025.11.003, pdfUrlCn=https://castjournals.cast.org.cn/joweb/wxdgc/CN/PDF/10.3969/j.issn.1003-3106.2025.11.003, pdfUrlEn=https://castjournals.cast.org.cn/joweb/wxdgc/EN/PDF/10.3969/j.issn.1003-3106.2025.11.003, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于无监督领域自适应的调制识别算法
收藏切换
PDF下载
昌硕 1 , 胥顺 1 , 魏梅英 2, *
无线电工程 | 信号与信息处理 2025,55(11): 2153-2162
收起
收藏切换
无线电工程 | 信号与信息处理 2025, 55(11): 2153-2162
基于无监督领域自适应的调制识别算法
全屏
昌硕1, 胥顺1, 魏梅英2, *
作者信息
  • 1.北京邮电大学 网络空间安全学院,北京 100876
  • 2.国家无线电监测中心,北京 100086
  • 昌硕 男,(1993—),博士,讲师,硕士生导师。主要研究方向:频谱深度认知,多源信息融合,深度学习。

    胥顺 男,(2000—),硕士研究生。主要研究方向:智能信号处理、自动调制识别、深度学习。

通讯作者:

魏梅英 女,(1985—),硕士,高级工程师。主要研究方向:无线电监测、无线电干扰定位、信号分析。
Modulation Recognition Algorithm Based on Unsupervised Domain Adaptation
Shuo CHANG1, Shun XU1, Meiying WEI2, *
Affiliations
  • 1.School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • 2.State Radio Regulation of China, Beijing 100086, China
出版时间: 2025-11-05 doi: 10.3969/j.issn.1003-3106.2025.11.003
文章导航
收藏切换

调制识别是无线通信中的关键任务,深度学习方法虽已取得显著进展,但在复杂非合作环境下仍面临泛化能力不足的挑战。特别是,变化的信道条件会模糊结构相似的调制方式(如16QAM与64QAM)之间细微的判别性特征,导致识别性能下降。针对这一调制识别领域的特有挑战,提出特征对齐对抗领域自适应(Feature Alignment and Discrimination Domain Adaptation,FADDA)的无监督对抗领域自适应算法。该方法的核心是在对抗训练的基础上,创新性地引入了基于对比学习的特征对齐损失。对抗训练负责学习域不变特征以适应信道变化,特征对齐损失则通过显式地增强类内特征的紧凑性和类间特征的可分性,从而在根本上提升模型对易混淆调制类型的分辨能力。实验验证表明,该方法在缺乏目标域标签的情况下,能够显著增强模型的跨信道调制识别性能,展现出强大的泛化能力。

调制识别  /  无监督学习  /  域自适应

Modulation recognition is a critical task in wireless communications. Although deep learning methods have achieved remarkable progress in this field, they still face the challenge of insufficient generalization ability in complex non-cooperative environments—particularly when confronted with varying channel conditions, which can obscure the subtle discriminative features between structurally similar modulation schemes (e. g. 16QAM and 64QAM) and thus degrade recognition performance. To address this unique challenge in the field of modulation recognition, an unsupervised adversarial domain adaptation method named Feature Alignment and Discrimination Domain Adaptation ( FADDA) is proposed. The core of FADDA is the introduction of a contrastive learning-based feature alignment loss on the basis of adversarial training. Adversarial training is responsible for learning domain-invariant features to adapt to channel variations, while the feature alignment. loss fundamentally enhances the model' s ability to distinguish between easily confused modulation types by explicitly reinforcing the compactness of intra-class features and the separability of inter-class features. Experimental results show that without target-domain labels, this method can significantly improve the model's cross-channel modulation recognition performance and demonstrate strong generalization ability.

modulation recognition  /  unsupervised learning  /  domain adaptation
昌硕, 胥顺, 魏梅英. 基于无监督领域自适应的调制识别算法. 无线电工程, 2025 , 55 (11) : 2153 -2162 . DOI: 10.3969/j.issn.1003-3106.2025.11.003
Shuo CHANG, Shun XU, Meiying WEI. Modulation Recognition Algorithm Based on Unsupervised Domain Adaptation[J]. Radio Engineering, 2025 , 55 (11) : 2153 -2162 . DOI: 10.3969/j.issn.1003-3106.2025.11.003
调制技术作为无线通信系统的核心组成部分,贯穿了通信技术从1G到5G乃至未来6G的发展历程,通过将基带信号转换为适合信道传输的带通信号,有效提升了通信系统的频谱利用率和抗干扰能力。随着智能城市、车联网、工业互联网等新兴应用的兴起,通信系统正朝着高速率、低时延、高可靠性方向演进,对调制识别的实时性与准确性提出了更高的要求[1-2]
调制识别技术旨在接收端在无需先验知识的情况下自动判断信号调制方式,是保障通信安全与效率的关键支撑能力。该技术广泛应用于频谱监测、干扰识别、电子侦察、认知无线电等场景。尤其在非合作通信环境中,调制识别是解调与信息还原的前置环节,其准确性直接关系到后续通信任务的可达性与稳定性[3-4]
传统调制识别方法,如基于似然判决和人工特征提取的方法,虽然理论完备、逻辑清晰,但存在计算复杂度高、对先验知识依赖强、泛化能力不足等问题,难以适应当前动态、复杂、多变的通信环境[5-8]。近年来,深度学习技术的引入为调制识别开辟了新路径,通过端到端学习自动提取信号特征,显著提升了识别精度和系统鲁棒性,成为当前调制识别研究的主流方向[9-11]
尽管如此,现有深度模型普遍依赖于固定场景下的标注数据,在面向不同信道条件、信噪比(Signal to Noise Ratio,SNR)水平、采样率或设备参数变化时,性能下降明显,泛化能力不足已成为调制识别落地应用的瓶颈。为此,本文围绕“提升深度学习调制识别模型在未知通信场景下的泛化能力”这一核心问题展开研究。然而,经典的对抗领域自适应方法(如领域对抗神经网络(Domain-Adversarial Neural Network,DANN))虽然旨在学习域不变特征,但其主要关注域级别的对齐,可能忽略了类别级别的判别结构。在调制识别任务中,这一问题尤为突出。例如,QPSK与8PSK,16QAM与64QAM等调制方式,其星座图结构和信号特征本身具有高度的相似性。在复杂的信道衰落和噪声干扰下,这些本就细微的差异变得更加模糊,仅依靠对抗训练可能不足以有效分离这些易混淆的类别,甚至可能在对齐域分布时反而拉近了它们的特征表示。
针对这一调制识别任务特有的挑战,本文提出一种结合了对抗训练与对比学习的FADDA算法。该方法的核心思想是:在通过对抗训练构建的域不变特征空间中,引入特征对齐损失,以一种有监督的方式(利用源域标签和目标域伪标签)显式地优化特征表示,目标是使同类调制的特征在空间上更紧凑,异类调制特征更分散。通过这种方式,本文方法不仅实现了跨域的泛化,更重要的是提升了模型在面对相似、易混淆调制信号时的判别精度。
自动调制识别(Automatic Modulation Recognition,AMR)是智能通信系统的关键技术之一。传统AMR方法主要分为两大类:基于最大似然(Maximum Likelihood,ML)的方法和基于特征提取(Feature-Based,FB)的方法。基于ML的方法在低SNR下能达到理论最优性能,但其计算复杂度极高且严重依赖精确的信道先验知识,在非合作通信场景下实用性受限。基于FB的方法通过专家手工设计特征(如高阶累积量、循环谱等)结合分类器进行识别[5-8],虽然降低了复杂度,但特征设计过程繁琐,且设计的特征在多变的信道环境下鲁棒性不足。
近年来,以深度学习为代表的人工智能技术[12-13],以其强大的端到端自动特征学习能力,为AMR领域带来了革命性的突破。研究者们成功地将多种深度神经网络应用于此任务。例如,Zhang等[3]开创性地使用卷积神经网络(Convolutional Neural Network,CNN)直接从原始I/Q信号样本中学习特征,并发布了公开数据集,极大地推动了该领域的研究进展。此后,研究者进一步探索了更复杂的网络结构,如利用循环神经网络(Recurrent Neural Network,RNN)或长短时记忆网络(Long Short-Term Memory,LSTM)来捕捉信号的时序相关性,以及将CNN与RNN相结合的混合模型,以充分利用信号的空时信息。这些研究的综述和挑战在文献[9]中有详细的讨论。然而,现有深度学习模型的一个根本局限在于其性能高度依赖训练数据与测试数据满足独立同分布(i. i. d.)的假设。在实际应用中,无线信道环境(如加性高斯白噪声(Additive White Gaussian Noise,AWGN)、瑞利衰落、莱斯衰落)的动态变化构成了典型的域偏移(Domain Shift)问题,导致在源信道环境(源域)训练好的模型,在新的未知信道环境(目标域)中性能会急剧下降[12],这正是本文致力于解决的核心挑战。
为解决域偏移问题,无监督领域自适应(Unsupervised Domain Adaptation,UDA)技术应运而生[14]。UDA旨在利用一个带有丰富标签的源域数据,提升模型在只有一个无标签目标域上的性能。其核心思想是学习一个对域间差异不敏感,但对主要任务(如分类)具有判别力的特征表示。根据实现方式的不同,主流UDA方法可分为两大类:
① 基于差异度量的自适应方法。此类方法通过显式地定义并最小化一个距离度量函数,拉近源域和目标域的特征分布。经典方法包括最小化最大均值差异(Maximum Mean Discrepancy,MMD)[15],以及深度相关对齐(Deep CORAL)[16]。例如,CORAL通过对齐源域与目标域特征分布的二阶统计量(即协方差矩阵)实现分布匹配。这类方法虽然概念清晰、计算高效,但主要关注分布的低阶矩,可能无法完全捕捉和对齐分布之间更复杂的高阶非线性结构,因此在面对复杂的信道畸变时效果可能有限。
② 基于对抗学习的自适应方法。此类方法借鉴了生成对抗网络(Generative Adversarial Network, GAN)的思想[17],通过引入一个域判别器与特征提取器进行“二人零和博弈”,隐式地学习域不变特征。DANN是该方向的开创性工作[12]。DANN的特征提取器旨在学习能够“欺骗”域判别器的特征,而域判别器则努力区分特征的来源域。通过这种对抗博弈,特征提取器被迫学习到无法区分域来源的、具有泛化性的特征表示。许多后续工作对DANN进行了改进,例如,条件领域对抗网络(Conditional Domain Adversarial Network,CDAN)指出,DANN仅对齐了边缘分布,可能导致属于不同类别的目标域样本被错误地映射到源域的某个类别特征簇附近。为解决此问题,CDAN[18]将类别预测信息与特征进行多线性映射后一同输入到域判别器中,使得对抗过程不仅考虑域信息,还考虑了类别信息,从而促使模型学习更具判别性的域不变特征。
然而,上述主流UDA方法大多致力于实现全局的域特征对齐,较少深入探讨任务本身的内在特性。在调制识别领域,一个独特的挑战是多种调制信号之间存在固有的高相似性,例如QPSK与8PSK, 16QAM与64QAM,它们在星座图结构、符号能量分布等方面非常接近。在这种情况下,单纯的对抗训练虽然能拉近域间距,但可能不足以分离开这些在特征空间中本就拥挤的类别,甚至可能为了“混淆”域判别器而牺牲了类别间的判别性边界。
为此,本文提出的FADDA算法旨在填补这一空白。算法没有停留在单一的对抗域适应框架,而是将对抗训练与对比学习的思想相结合。通过在对抗生成的域不变特征空间中,引入一个显式的特征对齐损失,强制性地拉近同类信号的特征、推远异类信号的特征。这种设计使得模型在实现对信道变化的鲁棒性的同时,能够更专注于增强对易混淆调制信号的判别能力,从而更适应调制识别任务的实际需求。
该算法框架主要由特征提取网络、分类器和领域判别器三部分组成,如图1所示。该算法是对经典领域对抗自适应算法[14]的改进,其核心思想是通过对抗训练对齐源域和目标域的特征分布,从而减少跨信道条件下的分布差异。同时,引入特征对齐损失,进一步增强同类别样本的紧凑性,并增大不同类别样本之间的区分度,从而提升模型在未知信道环境下的泛化能力。接下来,将详细介绍该方法的网络架构、损失函数设计以及训练过程。
特征提取器用于从输入信号中提取深度特征,具体结构如图2所示。特征提取器表示为Fx,θf),其中θf表示特征提取器中所有可训练的参数。
分类器(C)用于对提取的特征进行分类,由2层全连接层组成。输出为不同分类类别的logist值,分类器表示为Cx,θc),其中θc表示分类器中所有可训练的参数。分类器网络架构如图3所示。
判别器(D)用于区分源域和目标域的特征分布。在特征输入后,首先经过梯度反转层(Gradient Reversal Layer,GRL),该层在前向传播时保持特征不变,但在反向传播时反转梯度的方向。判别器由全连接层组成,用于对反转梯度后的特征进行领域分类。输出为不同领域类别的logist值。判别器表示为Dx,θd),θd表示判别器中所有可训练的参数。判别器网络架构如图4所示。
由于目标域数据是无标签的,因此仅对源域数据计算分类损失:
式中:表示分类器对源域特征的预测结果,ys表示源域真实的标签,CE表示交叉熵损失函数。最小化Lc表示分类器能够对源域特征进行正确分类。
为了实现特征对齐,使得类内样本更加紧凑,类间样本更加可分,采用对比损失来优化特征表示。通过计算样本间的相似度关系,引导相同类别的样本特征聚集,而不同类别的样本特征远离,从而提高分类的判别能力。
式中:fs表示源域特征,yt表示高置信度样本筛选后的目标域样本伪标签,ft表示高置信度样本筛选后的目标域样本特征。
CL是对比损失,其核心目标是通过衡量样本间的相似性,使得同类别样本的特征更加紧凑,而不同类别的样本特征更加可分。对比损失的形式如下:
式中:,采用归一化特征的余弦相似度衡量样本对之间的相似程度。Pi)={j|yi=yj,ji}表示所有与样本i具有相同标签的样本,Ni)={j|gigj}表示所有与样本i具有不同标签的样本,T表示温度系数,用于控制相似度分布的敏感性。
领域分类器用于区分源域与目标域特征,损失函数为:
式中:分别表示源域和目标域特征经过判别器得到的预测结果,dsdt表示源域和目标域的真实标签,通常,源域的标签为0,目标域的标签为1;CE表示交叉熵损失函数。最小化Ld表示判别器能够正确区分源域特征和目标域特征,通过在网络中引入梯度反转层,特征提取器的参数将朝着相反的方向进行优化,使得源域和目标域的特征变得不可区分,从而使得特征提取器学习到的特征更加鲁棒,不依赖于域的差异。
最终,综合以上3种损失函数,优化目标为:
式中:λ1λ2为超参数,用于调整不同损失项之间的权重,确保模型在源域分类性能、特征对齐和领域不变性之间取得平衡。本算法的优化目标(式(5))包含3个相互关联的部分。LcLf旨在优化特征的类别判别性,而Ld则通过与判别器D的min-max博弈来促进特征的域不变性。二者之间存在一种协同与制衡的关系。
协同机制:对抗损失(Ld)迫使特征提取器F学习一个“模糊”了域来源的特征空间。在这个域不变的特征空间基础上,特征对齐损失(Lf)才能更有效地发挥作用,因为它操作的特征已经消除了大部分与信道相关的干扰,使其能更专注于信号调制方式本身的差异。
平衡机制:超参数λ1λ2在此过程中扮演了平衡者的角色。λ2Ld旨在抹平特征的域差异,而λ1Lf则旨在强化特征的类别差异。若λ1设置过高,模型可能过度拟合源域的类别分布,从而破坏好不容易建立的域不变性,导致在目标域上性能下降。反之,若λ1过低,则不足以将目标域中易混淆的类别(如16QAM和64QAM)分离开。因此,通过调λ1λ2的相对大小,可以控制模型在“域不变性”和“类别判别性”之间的权衡。虽然对该复杂优化目标的严格收敛性证明超出了范畴,但通过实验调优找到了一个有效的平衡点,确保了训练的稳定性和最终的性能增益。
首先准备训练数据,记为,其中是输入样本,是对应的标签。目标域数据包含大量未标注数据,其中是输入样本,标签未知。同时,为每个样本分配一个领域标签,表示样本来自源域还是目标域,其中,, N=Ns+ Nt为样本总数且Ns=Nt
初始化网络参数特征提取器参数θf、分类器参数θd和判别器参数θd
每次训练从Ns个源域样本和Nt个目标域样本个中各采样一个批次Nbs。然后计算前向传播时的损失,根据分类器的预测结果计算分类损失为:
计算分类器在目标域的每个样本预测结果yt=,同时移除梯度,以避免参与后续反向传播。将其归一化后根据置信度阈值τ,挑选高置信度样本,并得到筛选后的目标域样本伪标签。以及筛选后的目标域样本,其数量为N′, 0≤N′Nbs,然后计算特征对齐损失:
最后,根据判别器的预测结果计算领域分类损失:
总损失为:
反向传播根据损失计算参数梯度,由于梯度反转层的存在, ∂Ld/θf需要取反。进行参数更新:
式中:μ表示学习率。
为研究不同信道条件下深度学习模型的泛化能力,通过GNU Radio工具箱的通信仿真模块构建了DataSet_AWGN、DataSet_Rayleigh和DataSet_Rician三个数据集。这些数据集在调制方式、SNR范围及采样参数等方面保持一致,仅信道条件不同,以便公平评估模型在不同信道环境中的适应性。数据集包含11种调制方式,包括OOK、4ASK、BPSK、QPSK、8PSK、16QAM、64QAM、16APSK、32APSK、GFSK、CPFSK。SNR设定为-6~18 dB,步长2 dB,即{-6,-4,-2,0,2,…,18},共13个SNR级别,涵盖从低SNR到高SNR的不同场景,训练集和测试集的比例为7:3。每个数据集的信号参数保持一致,具体配置如表1所示。
接下来介绍3个数据集的信道配置。
DataSet_Awgn:仅考虑AWGN,不包含多径衰落效应。适用于理想无线信道环境。
DataSet_Rayleigh:采用了Rayleigh衰落信道模型,仅考虑散射多径效应而不包含直射路径。该信道配置模拟了城市微蜂窝环境或室内热点场景,适用于低速移动或静态场景。Rayleigh信道参数配置如表2所示。
DataSet_Rician:采用Rician衰落信道模型,包含直射路径和散射多径效应。该信道配置模拟了室外开阔区域。农村田野或城市中视线无遮挡的开阔区域。Rayleigh信道参数配置如表3所示。
训练时首先加载预训练模型,确保模型从已有知识开始训练,提高训练效率,每批次输入的数据大小为512,训练总共进行200个周期。使用不同的优化器分别优化特征提取器,分类器以及判别器。其中特征提取器采用Adam优化器,初始学习率设置为0.0001。分类器也采用Adam优化器,它的初始学习率设置为0.005。判别器也采用Adam优化器,初始学习率设置为0.001。最后,根据最高准确率选择最佳模型。超参数设置方面特征对齐损失系数λ1设置为0.1,对比损失的温度T设置为0.1,伪标签置信度阈值τ设置为0.8,λ2选择经典DANN中的设置λ2=2/(1+exp(-10×p)),其中p为当前迭代次数与总迭代次数的比率,逐步增强领域对齐损失的影响。
该算法与不加对齐损失的基线算法DANN进行了对比分析。当不加对齐损失时,算法退化为DANN,仅通过对抗训练进行特征对齐。实验采用所有SNR下的平均识别精度作为评估标准,以全面衡量算法性能。此外,为了提供更清晰的性能评估基准,实验设置了模型性能的上界和下界。其中,上界指的是假设目标域数据带有真实标签时,直接在目标域上训练模型所能达到的最高精度,代表理想情况下的性能上限;下界则表示仅在源域数据上训练并直接在目标域上测试时的精度,反映无领域自适应时的性能下限。
各个算法分类性能如表4所示。
FADDA相比于下界(无领域自适应)均有明显提升,在3个迁移任务下,相对于下界的性能分别提升约23.8个百分点、32.6个百分点、35.3个百分点,平均提升约30.5个百分点。其中,在AWGN-Rayleigh和AWGN-Rician的迁移任务中,FADDA的识别精度距离上界仅相差4.9个百分点和8.4个百分点,这表明领域自适应算法能够有效减少源域和目标域之间的特征分布差异,从而提升模型在跨信道条件下的泛化能力。
相对于经典的DANN算法,FADDA通过引入对齐损失进一步提升了模型性能。3种任务下,FADDA均取得了优于DANN的表现,平均领先约1.56个百分点。这表明FADDA通过结合对齐损失和对抗训练,使得源域和目标域的特征分布更加一致,从而在跨信道条件下表现出更强的适应能力。
Rayleigh-Rician迁移难度最高,特征差异分布较大,下界精度仅有0.366,即使是FADDA方法,最终识别精度也仅达到0.719,距离上界还有19.9个百分点的差异,远低于其他迁移任务的表现。这一结果表明,在特征分布差异较大的场景中,该领域自适应算法的性能仍有较大提升空间。未来的研究可以进一步探索更高效的方法,以提升模型在跨域差异大的场景下的适应能力和识别精度。
为了进一步验证本文提出的FADDA模型中各个组件的有效性,设计了一组消融实验。以AWGN→Rayleigh这一代表性迁移任务为例,系统地分析了对抗损失(Ld)和特征对齐损失(Lf)各自以及共同的作用。实验设置包括:
① 下界:仅使用源域训练的基线模型,即L=Lc
② DANN:经典的对抗自适应网络,损失为L=Lc+λ2Ld
③ Ours(w/o Ld):移除了对抗损失,仅使用本文的特征对齐损失进行自适应,损失为L=Lc+λ1Lf
④ FADDA:损失为L=Lc+λ1Lf+λ2Ld
不同信道下迁移前后性能对比如表5所示。
表4表5可以看出:① 与下界相比,单独使用对抗损失(DANN)或特征对齐损失(Ours w/o Ld)都能显著提升模型的性能,证明了这2种自适应策略的独立有效性。② DANN的性能略优于仅使用特征对齐损失的模型,表明全局的域对齐在跨信道任务中起着至关重要的基础作用。③ FADDA取得了最佳性能,优于所有其他变体。这有力地证明了对抗损失和特征对齐损失之间存在着强大的互补性。对抗训练负责构建一个对信道变化鲁棒的、粗粒度的共享特征空间,而特征对齐损失则在此空间内进行细粒度的类别结构调整,确保了相似类别的可分离性。二者结合,最终实现了最优的跨域识别性能。
选择了一个具有代表性的信道迁移任务,即从AWGN信道到瑞利衰落信道的迁移任务,分析了不同SNR下领域自适应前后的分类精度变化。通过对比FADDA包括DANN的性能,以及模型的上界和下界,评估了它们在低SNR和高SNR条件下的表现。实验结果如图5所示。
随着SNR的增加,所有方法的分类精度均显著提升。从-6~18 dB的平均分类精度来看,FADDA的平均分类精度最高(达到86.5%),远高于理论下限的67.9%,提升幅度约为27个百分点。
在低SNR条件下(SNR≤0 dB),由于噪声干扰较强,所有方法的分类精度均较低,但FADDA依然显著优于下界,表现出较强的领域自适应能力。FADDA的平均分类精度分别为61.6%,相比下界43.5%提升了18.1个百分点,表明在极端噪声条件下,FADDA能够有效提取迁移特征,减少跨信道分布差异。而DANN的分类精度在同样的SNR下为58.0%,平均比FADDA略低3.6个百分点。
在中等SNR条件下(2 dB≤SNR≤8 dB),FADDA的分类精度进一步提升,相较于下界的优势更加明显。FADDA在该SNR范围内的平均精度为91.5%,相比下界66.2%提升了25.3个百分点,表明在中等SNR条件下,领域自适应方法的作用尤为显著。然而,在该SNR范围内,FADDA与DANN的分类精度差距较小,DANN的平均精度为90.8%,仅比FADDA低0.7个百分点,说明在该SNR范围内, FADDA的特征对齐机制对性能提升的作用相对有限。
在高SNR条件下(SNR≥10 dB),FADDA分类精度基本收敛,接近上界,表明其在高SNR下具有更强的稳定性。DANN在该SNR范围内的平均精度比FADDA低2.2个百分点,说明即使在高SNR条件下,FADDA仍能通过特征对齐进一步优化识别效果,使得特征更加紧凑,提高模型的最终分类性能。
本小节将分析AWGN→Rayleigh任务下,基于FADDA算法,针对不同调制信号在特定SNR条件下领域自适应前后的分类情况。图6图7分别展示了在0 dB和8 dB SNR条件下,领域自适应前后的混淆矩阵。混淆矩阵的横坐标为模型的预测结果,纵坐标为样本信号的真实类别。图中的颜色深浅代表了分类概率的大小,颜色越深表示分类概率越高。混淆矩阵的数字,代表某一类别的分类准确率。
自适应前分析:在没有进行领域自适应时,模型表现出严重的性能下降。值得注意的是,其误分类并非随机的,而是呈现出与信道特性高度相关的模式。例如,OOK和4ASK作为纯幅度调制信号,在瑞利衰落信道中其幅度信息被剧烈且随机地改变,导致其信号特征与本身就包含复杂幅相信息的APSK类信号(16APSK, 32APSK)高度混淆。同样,BPSK约有一半被误判为16PSK,这也是衰落信道模糊了其简单的相位特征所致。这表明模型将在源域(AWGN)学到的、依赖于稳定幅度和相位的判别依据,在目标域(Rayleigh)完全失效。
自适应后分析:经过FADDA自适应后,分类性能得到显著改善。最明显的变化是,由信道特性引起的跨族系(如ASK vs APSK)的严重误分类现象得到了极大遏制。例如,OOK和4ASK的识别准确率从几乎为0提升到了70%和91.5%。证明了本文方法成功学习到了对瑞利衰落信道鲁棒的特征表示。然而,部分混淆依然存在,主要集中在结构相似的类别内部,例如64QAM有大量样本被误判为16QAM,32APSK有不少被误判为16APSK。这揭示了在低SNR下,即使域差异被对齐,高阶与低阶调制方式之间的内在相似性仍然是识别的瓶颈。总体而言,FADDA显著提升了模型的稳定性和准确性,将主要的识别困难从“无法适应信道”转变为“分辨高阶信号”这一更根本性的问题。
图6中0 dB的情况相比,混淆矩阵的对角线更加清晰,表明在高SNR下分类性能显著提升。图7(a)显示,在领域自适应前,主要的分类误差来源于OOK、4ASK和BPSK,这些信号大多被误判为16APSK和32APSK。OOK、BPSK和4ASK的识别主要依赖单一的幅度信息,而在瑞利衰落信道下,信号幅度会发生随机变化,导致幅度特征模糊,幅度特征与APSK的幅度分布相似,从而引发误分类。此外,还存在少量误差,主要来自于QAM和APSK类内的差异。图7(b)表明,经过领域自适应后,误差主要集中在16QAM和64QAM之间的误判。由于16QAM可以被视为64QAM的子类,因此这种误差是可以理解的。总体而言,领域自适应显著提升了分类的准确性,减少了误判情况。
本文成功构建并验证了一种名为FADDA的UDA调制识别算法,其核心贡献在于揭示了仅依靠全局域对抗训练在应对信道变化,尤其是在区分星座图结构相似的调制类型时存在的局限性。实验清晰地证实,本文的关键创新——在对抗训练构建的域不变特征空间中,进一步引入基于对比学习的特征对齐损失——是提升模型性能的核心。该设计超越了传统方法,通过显式优化特征空间,增强了同类调制的“内聚性”和异类调制的“可分性”,从而成功将识别瓶颈从宏观的“无法适应信道”问题,转化为微观的“如何分辨高阶相似信号”问题,这标志着模型在本质上学习到了对信道变化更具鲁棒性的深层特征。因此,本研究的价值不仅是提供了一个性能更优的算法,更在于提出了一种将全局领域对齐与局部类别结构优化相结合的新思路,为解决其他信号处理领域中面临“域偏移”与“细粒度分类”双重挑战的复杂场景提供了有价值的参考,并为开发下一代高泛化能力的智能信号识别系统奠定了坚实的基础。
参考文献 引证文献
排序方式:
[1]
SAAD W, BENNIS M, CHEN M Z. A Vision of 6G Wireless Systems: Applications, Trends, Technologies, and Open Research Problems[J]. IEEE Network,2019, 34(3):134-142.
[2]
LETAIEF K B, CHEN W, SHI Y M, et al. The Roadmap to 6G: AI Empowered Wireless Networks[J]. IEEE Communications Magazine,2019,57(8):84-90.
[3]
ZHANG F X, LUO C B, XU J L, et al. Deep Learning Based Automatic Modulation Recognition: Models, Datasets, and Challenges[J]. Digital Signal Processing,2022, 129:103650.
[4]
WANG Y,YANG J,LIU M,et al. LightAMC: Lightweight Automatic Modulation Classification via Deep Learning and Compressive Sensing[J]. IEEE Transactions on Vehicular Technology,2020,69(3):3491-3495.
[5]
HO K C, PROKOPIW W, CHAN Y T. Modulation Identification by the Wavelet Transform[C]//Proceedings of MILCOM'95. San Diego:IEEE,1995:886-890.
[6]
张文娜,孙运强,姚爱琴,.基于高阶累积量的数字调制识别[J].国外电子测量技术,2022,41(7):126-132.
[7]
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[J].电子与信息学报, 2016,38(3):674-680.
[8]
孔豫京,黄焱,马金全. Alpha稳定分布噪声背景下衰落信号的调制识别方法[J].信号处理,2015,31(8):975-985.
[9]
ZHOU R L, LIU F G, GRAVELLE C W. Deep Learning for Modulation Recognition: A Survey with a Demonstration[J]. IEEE Access,2020,8:67366-67376.
[10]
KUMAR Y, JAJOO G, YADAV S K. 2D-FFT Based Modulation Classification Using Deep Convolution Neural Network[C]//2020 IEEE 17th India Council International Conference (INDICON). New Delhi:IEEE,2020:1-6.
[11]
ZHANG R, HE C B, JING L Y, et al. A Modulation Recognition System for Underwater Acoustic Communication Signals Based on Higher-order Cumulants and Deep Learning[J]. Journal of Marine Science and Engineering,2023,11(8):1632.
[12]
董聪,张传武,高勇.基于残差神经网络的通信混合信号识别[J].无线电工程,2020,50(9):727-731.
[13]
杨梓辰,笪清安,李思照.面向深度学习模型的对抗攻击与防御综述[J].无线电工程,2023,53(6):1328-1341.
[14]
LIANG Z, XIE J, YANG X, et al. Self-training Based Adversarial Domain Adaptation for Radio Signal Recognition[J].IEEE Communications Letters,2022,26(11):2646-2650.
[15]
ARBEL M, KORBA A, SALIM A, et al. Maximum Mean Discrepancy Gradient Flow[EB/OL]. (2019-06-11)[2025-05-11]. https://arxiv.org/abs/1906.04370.
[16]
SUN B C,SAENKO K. Deep CORAL:Correlation Alignment for Deep Domain Adaptation[C]//European Conference on Computer Vision. Amsterdam:Springer,2016:443-450.
[17]
GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative Adversarial Nets[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems. Montreal:ACM,2014:2672-2680.
[18]
LONG M S, CAO Z J, WANG J M, et al. Conditional Adversarial Domain Adaptation[EB/OL]. (2017-05-26)[2025-05-11]. https://arxiv.org/abs/1705.10667.
2025年第55卷第11期
PDF下载
141
67
引用本文
BibTeX
文章信息
doi: 10.3969/j.issn.1003-3106.2025.11.003
  • 接收时间:2025-07-07
  • 首发时间:2026-04-17
  • 出版时间:2025-11-05
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-07-07
基金
作者信息
    1.北京邮电大学 网络空间安全学院,北京 100876
    2.国家无线电监测中心,北京 100086

通讯作者:

魏梅英 女,(1985—),硕士,高级工程师。主要研究方向:无线电监测、无线电干扰定位、信号分析。
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/wxdgc/CN/10.3969/j.issn.1003-3106.2025.11.003
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏