SIA OpenIR  > 机器人学研究室
基于sEMG的手部康复机器人交互控制研究
Alternative TitleResearch on Interactive Control of Hand Rehabilitation Robot Based on sEMG
马乐乐1,2
Department机器人学研究室
Thesis Advisor赵新刚
Keyword康复机器人 表面肌电信号 特征提取 BP神经网络 卷积神经网络
Pages70页
Degree Discipline模式识别与智能系统
Degree Name硕士
2019-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract脑卒中是急性的脑血管疾病,近年来发病率和致残率逐渐增高。医学研究表明,有效的康复训练能够刺激患者神经功能的恢复,从而缓解患者肢体运动的障碍并加快患者康复治疗的进程。由于传统训练治疗方式的各种弊端,患者、家庭、医院和康复治疗中心对手部康复机器人的需求量大幅度增加。现有的手部康复系统大多数是简单的被动训练模式,患者无法进行主动训练,因此在本课题中引入肌电信号。本文手部康复机器人通过采集患者的肌电信号来识别动作意图,并将其反馈给康复机器人用于辅助训练。本文的主要内容如下:首先,介绍了肌电信号的产生原理和肌电信号的特点。根据肌肉解剖位置粘贴电极,按顺序采集各个动作下的肌电信号。利用数字陷波器和巴特沃兹滤波器对肌电信号进行滤波去噪并利用最大面积法修正类别标签。然后利用时间窗和增量窗对每个通道的肌电信号进行分割并提取RMS、MAV和ARC等11个特征,对特征归一化后再利用PCA完成降维操作。其次,利用提取到的特征建立BP神经网络模型用于识别患者的动作类型。提出基于GBDT算法的肌电信号最优通道选择方法用于提高肌电识别系统的使用效率和鲁棒性。再次,建立基于多流卷积操作和全局池化的深度学习模型,用于对分割出的肌电图像进行特征自动提取,完成端到端的动作识别。在动作识别率和训练过程稳定性方面,与传统的RF算法、SVM算法和其他卷积网络结构进行了对比。验证了卷积神经网络模型能够提高基于肌电信号的动作识别率。最后,搭建了手部康复机器人的软硬件系统,实现了手部康复机器人的阈值控制和实时控制。并根据患者的不同病情设计了被动训练、镜像主动训练、患侧主动训练等不同的康复训练策略。同时,康复系统还引入了游戏辅助训练,可以提高患者训练过程的积极性。
Other AbstractStroke is an acute cerebrovascular disease, and the incidence and disability rate have gradually increased in recent years. Medical research has shown that effective rehabilitation training can stimulate the recovery of neurological function in patients, thereby alleviating the obstacles of limb movement and accelerating the progress of rehabilitation. Due to various drawbacks of traditional training and treatment methods, the demand for hand rehabilitation robots in patients, families, hospitals and rehabilitation centers has increased significantly. Since most of the existing hand rehabilitation systems are simple passive training mode, and patients cannot carry out active training, sEMG is introduced in this topic. In this paper, the hand rehabilitation robot recognizes the movement intention by collecting the patient's sEMG, and feeds it back to the rehabilitation robot to assist training. The main contents of this paper are as follows: First, the principle of the generation of myoelectric signals and the characteristics of myoelectric signals are introduced. The electrodes are attached according to the position of the muscle anatomy, and the myoelectric signals under each action are sequentially collected. The EMG signal is filtered and denoised using a digital notch and a Butterworth filter and the category label is corrected using the maximum area method. Then, using the time window and the incremental window, the EMG signals of each channel are segmented and 11 features such as RMS, MAV and ARC are extracted. After the features are normalized, the PCA is used to complete the dimensionality reduction operation. Secondly, the BP neural network model is built using the extracted features to identify the type of motion of the patient. An optimal channel selection method for EMG signals based on GBDT algorithm is proposed to improve the efficiency and robustness of EMG recognition system. Thirdly, a deep learning model based on multi-stream convolution operation and global pooling is established to extract features automatically from segmented EMG images and complete end-to-end movement recognition. In terms of recognition rate of movements and stability of training process, it is compared with traditional RF algorithm, SVM algorithm and other convolutional network structures. It is verified that the convolution neural network model can improve the recognition rate of movement based on sEMG. Finally, the software and hardware system of the hand rehabilitation robot was built, and the threshold control and real-time control of the hand rehabilitation robot were realized. According to the different conditions of patients, different rehabilitation training strategies such as passive training, active mirror training and active training on the affected side were designed. At the same time, the rehabilitation system also introduces game-assisted training, which can improve the enthusiasm of patients in the training process.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25170
Collection机器人学研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
马乐乐. 基于sEMG的手部康复机器人交互控制研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
基于sEMG的手部康复机器人交互控制研究(4362KB)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
[马乐乐]'s Articles
Baidu academic
Similar articles in Baidu academic
[马乐乐]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[马乐乐]'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.