2025年04月03日 星期四 首页   |    期刊介绍   |    编 委 会   |    投稿指南   |    期刊订阅   |    统合信息   |    联系我们
计量学报  2021, Vol. 42 Issue (5): 629-637    DOI: 10.3969/j.issn.1000-1158.2021.05.14
  无线电、时间频率计量 本期目录 | 过刊浏览 | 高级检索 |
基于混合迁移学习的运动想象分类算法研究及其在脑机接口中的应用
杜义浩1,刘兆军1,付子豪1,张园园1,任娜2,陈杰3,谢平1
1. 燕山大学河 北省测试计量技术及仪器重点实验室,河北 秦皇岛 066004
2. 燕山大学电气工程学院,河北 秦皇岛 066004
3. 燕山大学 体育学院,河北 秦皇岛 066004
Motion Imagery Classification Algorithm Research Based on Hybrid Transfer Learning and Application in Brain-computer Interface
DU Yi-hao1,LIU Zhao-jun1,FU Zi-hao1,ZHANG Yuan-yuan1,REN Na2,CHEN Jie3,XIE Ping1
1. Key Laboratory of Measurement Technology and Instrumentation of Hebei Province, Yanshan University,  Qinhuangdao,Hebei 066004, China 
2. Institute of Electrical Engineering, Yanshan University, Qinghuangdao, Hebei 066004, China
3. Institute of Physical Education, Yanshan University, Qinghuangdao, Hebei 066004, China
全文: PDF (2639 KB)   HTML (1 KB) 
输出: BibTeX | EndNote (RIS)      
摘要 为提升迁移学习在运动想象脑机接口应用过程中的迁移高效性及普适性,综合实例迁移和特征迁移学习方法的优势进而构建了混合迁移学习模型。首先,依据样本权重极化原理改进TrAdaBoost算法以实现实例层面的迁移,优化源域训练样本;其次,基于大间隔投射迁移支持向量机进一步缩短源域与目标域间的分布距离以完成特征层面的迁移,实现迁移效率最大化。进一步,将该方法应用于脑机接口竞赛Dataset IIb数据集进行离线测试及分析,研究结果表明混合迁移学习模型的迁移效率明显高于单一迁移学习模型,并且对于不同迁移对象识别准确率相对提升均值在70%以上,验证了所述方法的有效性与普适性。此外,基于已搭建的运动想象识别系统进行在线测试,验证了模型的实用性。
服务
把本文推荐给朋友
加入我的书架
加入引用管理器
E-mail Alert
RSS
作者相关文章
杜义浩
刘兆军
付子豪
张园园
任娜
陈杰
谢平
关键词 计量学运动想象脑机接口实例迁移特征迁移混合迁移    
Abstract:To improve the efficiency and universality of transfer learning in the application of motor imagery brain-computer interface (MI-BCI),a hybrid transfer learning model with integrating the advantages of instance transfer and feature transfer learning methods was built.Firstly,the transfer of the instance level by introducing the principle of sample weight polarization to improve the classical TrAdaBoost algorithm was realized,which can optimize training samples in the source domain to some extent.Secondly,to further narrow the distance between the source domain and the target domain,the large margin projected transductive support vector machine was applied to complete the transfer of the feature level,thus maximizing the transfer efficiency.Furthermore,the proposed method was applied to the BCI competition dataset (Dataset IIb data set) for offline test and analysis.The results showed that the hybrid transfer learning model achieved significantly better transfer efficiency than the single transfer learning model,and obtained an improvement of the recognition rate with average value of above 70% for different transfer objects.The results also verified the effectiveness and universality of the hybrid transfer learning model.In addition,the online test was carried out based on the established motor-imagery system,which further verified the practicability of the model.
Key wordsmetrology    motor imagery    brain-computer interface    instance transfer    feature transfer    hybrid transfer
收稿日期: 2019-09-27      发布日期: 2021-05-24
PACS:  TB973  
  TB99  
基金资助:国家自然科学基金(61673336);河北省自然科学基金(F2015203372);河北省高等学校科学技术研究项目(QN2016094)
通讯作者: 谢平(1972-),黑龙江齐齐哈尔人,燕山大学教授,主要从事脑机接口、智能康复等方面的研究。Email:pingx@ysu.edu.cn     E-mail: pingx@ysu.edu.cn
作者简介: 杜义浩(1983-), 河北沧州人,燕山大学讲师,主要从事康复机器人生物反馈控制等方面的研究。Email: duyihao@126.com
引用本文:   
杜义浩,刘兆军,付子豪,张园园,任娜,陈杰,谢平. 基于混合迁移学习的运动想象分类算法研究及其在脑机接口中的应用[J]. 计量学报, 2021, 42(5): 629-637.
DU Yi-hao,LIU Zhao-jun,FU Zi-hao,ZHANG Yuan-yuan,REN Na,CHEN Jie,XIE Ping. Motion Imagery Classification Algorithm Research Based on Hybrid Transfer Learning and Application in Brain-computer Interface. Acta Metrologica Sinica, 2021, 42(5): 629-637.
链接本文:  
http://jlxb.china-csm.org:81/Jwk_jlxb/CN/10.3969/j.issn.1000-1158.2021.05.14     或     http://jlxb.china-csm.org:81/Jwk_jlxb/CN/Y2021/V42/I5/629
京ICP备:14006989号-1
版权所有 © 《计量学报》编辑部
地址:北三环东路18号(北京1413信箱)  邮编:100029 电话:(010)64271480
本系统由北京玛格泰克科技发展有限公司设计开发  技术支持:support@magtech.com.cn