网站首页            期刊简介             编委会             投稿指南             期刊订阅             下载中心             在线留言            联系我们             English
  2025年4月30日 星期三  
文章快速检索
中国生物医学工程学报  2016, Vol. 35 Issue (5): 526-532    DOI: 10.3969/j.issn.0258-8021.2016.05.003
  论著 本期目录 | 过刊浏览 | 高级检索 |
基于双线性模型的动作肌电信号用户无关识别研究
成娟#, 陈勋#*, 彭虎
合肥工业大学仪器科学与光电工程学院生物医学工程系, 合肥 230009
Research on User-Independent Gesture Recognition Based on Bilinear Models for sEMG Signals
Cheng Juan# Chen Xun#* Peng Hu
Department of Biomedical Engineering,School of Instrument Science and Opto\|electronics Engineering Hefei University of Technology, Hefei 230009, China
全文: PDF (0 KB)   HTML (1 KB) 
输出: BibTeX | EndNote (RIS)      
摘要 动作肌电信号具有个体差异性且不同动作的肌电信号是不同的,通过挖掘双线性模型的因素分解能力,将训练样本的特征矢量分解为用户相关和动作相关两大因素,通过确定因素的维度重构具有共性的训练样本特征。在测试样本特征重构阶段引入适应融合机制,更新模型参数重构测试样本特征。以11名受试者的4类动作为例,分别采用线性判别、K近邻分类算法和支持向量机,对比3种实验方案(多用户单天、单用户多天和基于双线性模型的多用户单天)的识别结果。实验表明,双线性模型的平均识别率最低为85%以上,相比于单纯的多用户单天识别结果(平均识别率不高于75%)有显著提高(P <0.001),且相比于单用户多天的识别结果(平均识别率90%以上)差异性不显著(P >0.24)。双线性模型为基于动作识别技术的非特定人肌电控制系统提供了交互方案,且该模型具备将多用户单天的数据看成单用户多天数据的能力,提供了用户训练负担降低的可行性。
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
关键词 肌电控制手势识别特征提取双线性模型    
Abstract:Due to the fact that surface electromyography (sEMG) signals of the same gesture vary from different individuals (user-related) and various gestures produce different sEMG signals (motion-related), the sEMG signals can be treated as the interaction of the two factors. This study utilized bilinear models to extract user-independent features. We first factorized original training features into two factors, and the determination of the factor dimensions can help the reconstructed features have the maximum similarity. When original testing features from a novel user were available, they were used to adapt the two factors with the aid of the aforementioned model parameters and the reconstructed testing features by using the adapted factors were finally sent to the trained classifier for recognition. Eleven subjects were recruited with each performing 4 types of gestures. Three classifiers (linear discriminant classifier, K-nearest neighbor and support vector machine) were employed for the classification of the three tasks, termed as user-dependent cross-time (UDC), original-user-independent (OUI) and bilinear-models-user-independent (BMUI). Experimental results showed that the averaged classification accuracy of BMUI was at least 85% whereas that of OUI was not higher than 75%. The one-way ANOVA analysis demonstrated the significant improvement of BMUI (P<0.001). Besides, although the averaged accuracy of UDC was above 90%, higher than that of BMUI, they were statistically insignificant (P>0.24). The proposed method provided a practical solution to the interaction implementation of myoelectric control system based on gesture recognition techniques, and the training samples could be significantly reduced since each subject will conduct only once experiment for training.
Key wordsmyoelectric control    gesture recognition    feature extraction    bilinear models
收稿日期: 2016-01-07     
PACS:  R318  
基金资助:国家自然科学基金(61401138;81571760;61501164); 中国生物医学工程学会高级会员(Senior member, Chinese Society of Biomedical Engineering)
引用本文:   
成娟, 陈勋, 彭虎. 基于双线性模型的动作肌电信号用户无关识别研究[J]. 中国生物医学工程学报, 2016, 35(5): 526-532.
Cheng Juan Chen Xun Peng Hu. Research on User-Independent Gesture Recognition Based on Bilinear Models for sEMG Signals. Chinese Journal of Biomedical Engineering, 2016, 35(5): 526-532.
链接本文:  
http://cjbme.csbme.org/CN/10.3969/j.issn.0258-8021.2016.05.003     或     http://cjbme.csbme.org/CN/Y2016/V35/I5/526
版权所有 © 2015 《中国生物医学工程学报》编辑部
本系统由北京玛格泰克科技发展有限公司设计开发