SIA OpenIR  > 机器人学研究室
基于肌电的手部康复机器人智能交互技术研究
Alternative TitleResearch on Intelligent Interaction technology of Hand Rehabilitation Robot Based on EMG
王丰焱
Department机器人学研究室
Thesis Advisor赵新刚
Keyword手部康复机器人 肌电信号 Brunnstrom等级 深度学习
Pages92页
Degree Discipline模式识别与智能系统
Degree Name硕士
2020-05-26
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract脑卒中疾病是导致成年人残疾的主要原因,给家庭和社会造成了巨大影响,积极主动的康复训练有助于脑卒中患者康复。手作为人体最重要的运动和感觉器官之一,其运动模式复杂。同时,手也是身体末端器官,康复难度极大。因此,脑卒中患者手部康复极其重要。针对传统手部康复治疗方式存在康复训练成本高、康复时间不灵活和康复等级评估主观性强等弊端,本课题提出了基于表面肌电的患者运动意图识别和康复等级评估技术,并基于此开发了具有实时康复等级信息反馈和主被动康复训练方式的手部康复机器人系统。本文主要内容如下:首先,介绍了表面肌电信号的产生原理和特点。根据脑卒中患者的病症,确定了患者表面肌电数据采集实验范式,并对采集到的表面肌电信号进行归一化和滤波等预处理操作以提高信号信噪比。然后利用滑动窗技术对每个通道数据进行处理得到样本,并针对得到的样本提取常用特征。最后,提出了基于tsfresh特征自动提取工具的肌电特征提取方法。其次,针对患者运动意图识别精度低的问题,提出了能够适用于不同Brunnstrom等级患者的动作识别方法。针对多类别精细动作识别率低的问题,提出了动态多路时空卷积网络框架,可以同时利用肌电信号的时间和空间信息,从而提高了动作分类模型精度。针对非理想情况下基于肌电动作分类模型性能下降问题,提出了基于纹理特征和集成学习方法的环状传感器位置偏移校正方法。再次,提出了基于肌电的Brunnstrom等级自动评估方法。首先,确定了9种康复训练动作,并根据动作和康复等级之间的相关性筛选出了适用于康复评估动作。然后,将筛选动作应用于康复等级自动评估算法,从而提高等级评估精度。接着,针对不同等级患者确定了相应的康复训练方案。最后,基于康复等级自动评估技术提出了两种家庭式手部康复机器人系统方案。最后,搭建了具有实时康复等级信息反馈和主被动康复训练方式的手部康复机器人系统,完成了肌电传感器、电源电路、主控电路、控制程序、运动意图识别算法、康复等级评估算法和用户界面软件等设计与实现。
Other AbstractStroke is the main cause of adult disability, which has a great impact on the family and society. Active rehabilitation training is helpful for the rehabilitation of stroke patients. As one of the most important moving and sensory organs of human body, the movement mode of hand is complex. At the same time, the hand is also the end organ of the body, so it is very difficult to recover. Therefore, the hand rehabilitation of stroke patients is extremely important. In view of the disadvantages of traditional hand rehabilitation treatment, such as high cost of rehabilitation training, inflexible rehabilitation time and strong subjectivity of rehabilitation grade evaluation, this paper designs a hand rehabilitation robot with real-time rehabilitation grade evaluation feedback and active passive rehabilitation training mode by combining the technology of patients' motion intention recognition based on sEMG and the assessment technology of rehabilitation grade based on sEMG. The main research works of this paper are as follows: Firstly, the generating principle and characteristics of surface myoelectric signal are introduced. According to the symptoms of stroke patients, the experimental paradigm of myoelectric data collection is determined, and the acquired sEMG signals were normalized and filtered to improve the signal-to-noise ratio. Then, the sliding window is used to process the data of each channel to obtain samples, and the common features are extracted for these samples. Finally, a method to extract myoelectric signals features by tsfresh feature extraction tool is proposed. Secondly, in order to solve the problem of low accuracy of motion intention recognition for patients, a motion recognition method suitable for patients with different Brunnstrom grades is proposed. Aiming at the problem of low recognition rate of multi-category fine motions, a multi-channel dynamic spatiotemporal convolution network is proposed, which can simultaneously use the time and space information of myoelectric signals to improve the accuracy of the action classification model. Aiming at the problem of performance degradation of myoelectric motion classification model under non-ideal situations, a sensor position offset correction method based on texture features and integrated learning model is first proposed. Thirdly, an automatic evaluation method of Brunnstrom grade based on myoelectric signals is proposed. First, nine types of rehabilitation training movements are identified, and the movements suitable for rehabilitation evaluation are screened based on the correlation between the movements and the rehabilitation level. Then, the screening movements are applied to the rehabilitation level automatic evaluation algorithm, so as to improve the accuracy of the level evaluation. After that the corresponding rehabilitation training programs are determined for different levels of patients. Finally, two kinds of home-based hand rehabilitation robot system schemes are proposed based on the rehabilitation level automatic assessment technology. Finally, a hand rehabilitation robot system with real-time rehabilitation level evaluation feedback and active and passive rehabilitation training methods is built. The design and implementation of electromyography sensor, power circuit, main control circuit, control program, motion intention recognition algorithm, rehabilitation level evaluation algorithm and user interface program are completed.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/27124
Collection机器人学研究室
Affiliation中国科学院沈阳自动化研究所
Recommended Citation
GB/T 7714
王丰焱. 基于肌电的手部康复机器人智能交互技术研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2020.
Files in This Item:
File Name/Size DocType Version Access License
基于肌电的手部康复机器人智能交互技术研究(6579KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[王丰焱]'s Articles
Baidu academic
Similar articles in Baidu academic
[王丰焱]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[王丰焱]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.