中国科学院沈阳自动化研究所机构知识库
Advanced  
SIA OpenIR  > 机器人学研究室  > 学位论文
题名: 肌电信号分析方法研究及其在康复领域的应用
其他题名: The Research on EMG Analysis and its Application in Rehabilitation
作者: 熊安斌
导师: 刘光军 ; 韩建达
关键词: 表面肌电信号 ; 肌电信号分解 ; 特征提取 ; 动作分类 ; 康复疗效量化评估
索取号: R318.04/X68/2015
页码: 119页
学位专业: 模式识别与智能系统
学位类别: 博士
答辩日期: 2015-05-26
授予单位: 中国科学院沈阳自动化研究所
学位授予地点: 中国科学院沈阳自动化研究所
作者部门: 机器人学研究室
中文摘要: 表面肌电信号(Surface Electromyography, sEMG)是肌肉收缩时,运动单元动作电位序列(Motor Unit Action Potential Train, MUAPT)在检测电极处叠加形成的混合信号。它能够准确的反映人体的运动意图。而且,通过分析运动单元(Motor Unit, MU)的募集和发放信息以及MUAP的波形特征,有助于分析肌肉及运动神经系统功能状态。因此,sEMG已经在人机交互、假肢控制、病理分析上有广泛的应用。另一方面,人类逐渐步入老龄化社会,由于关节炎、帕金森、脑卒中等病症导致的行动不便患者逐年增多;此外,全世界每年大约500万人因为车祸、地震等不可抗拒因素导致永久性的肢体残疾;因此,越来越多的人需要接受物理康复训练,以重获肢体运动能力。而现有的物理康复治疗,主要是医生与患者之间一对一的康复方式,这导致了康复医生人数的急剧短缺。康复机器人技术的出现,为缓解这一矛盾提供了很好的解决方案,现已成为医疗机器人领域的研究热点之一。经过几十年的发展,康复机器人技术已经有了长足的进步,但仍然存在诸多问题:1) 现有的康复机器人大多按照机电一体化思路设计,不能感知人的运动意图;2) 少数基于生物电信号控制的康复机器人,需要传感器数量较多,且针对每个用户需要单独训练信号识别模型,康复效率低下;3) 现有康复疗效评估方法,往往依赖医生与患者的主观感受,缺乏实时量化评估方法,影响患者参与康复训练的积极性。本论文针对上述问题展开深入研究,主要包括面向康复机器人的人体运动意图感知方法和基于sEMG的康复疗效量化评估方法。具体内容如下:首先,现有的基于sEMG的康复设备控制方法,往往采集多块肌肉上的sEMG信号,通过其不同的模式组合,识别用户的运动意图;但是,针对手、面部等小肌肉群,由于其区域面积有限,而sEMG电极体积往往相对较大,不适合使用多通道sEMG。因此,本文提出一种基于单通道sEMG分解的手部动作识别方法。对sEMG信号进行阈值计算、尖峰检测、高斯混合模型 (Gaussian Mixture Model, GMM)聚类,从而得到多个通道的MUAPT;并在此基础上进行特征提取、特征降维、线性判别分析(Linear Discriminate Analysis, LDA)分类,从而识别出不同的手部动作。实验结果表明,本文提出的方法,准确率达到86.3%,而传统方法单通道sEMG识别方法,精度仅在76%左右。然后,由于每个用户肌肉形态、脂肪含量的差异,传统的sEMG识别模型,往往只适用于用户本人。因此,我们提出一种基于sEMG分解的与个体无关(User-Independent)的sEMG识别方法。首先,采用分层聚类方法对检测到的sEMG尖峰进行聚类,以类别中心作为模板;然后,采用Gram-Schmidt正交法对模板空间进行优化,并将sEMG投影到该高维空间中;最后,利用深信念网络进行分类,从而识别出不同的手部动作。本文针对于不同用户的sEMG信号建立了一个通用的分类模型。实验结果表明,该方法的动作识别准确率达到81.5%,而传统的sEMG识别方法,准确率为70%左右.第三,传统的康复疗效评价方法,主要依赖医生的临床经验以及患者的主观感受,缺乏客观准确的评价标准。本研究在充分分析sEMG的单个时域、频域特征的基础上,提出利用sEMG的混沌特征,对康复疗效进行评估的方法。首先,利用时间延迟法重构sEMG的多维相空间;随后,计算关联维数与李雅普诺夫指数,验证sEMG的混沌特性;最后,通过近似熵量化sEMG的不确定性。此外,本研究采集了21名面瘫患者的面部双侧咬肌、提上唇肌和额肌处的sEMG信号,对上述分析方法进行验证。结果表明,面瘫患者健康侧与患侧sEMG的混沌特征表现出显著的统计学差异,sEMG的混沌特征能够作为面瘫诊断与康复评价的参考指标。最后,针对上述的sEMG混沌特征只是定性的区分健康侧与患侧的sEMG,而不能对康复疗效进行量化评估的问题,本研究将分别采用基于多通道sEMG特征组合和单通道sEMG分解的方法,量化康复疗效并预测患者康复趋势。前者对健康侧与患侧的多通道sEMG分别提取时频域特征,然后采用k均值聚类方法得到类别中心距离,并构建自回归模型(Auto Regression Model, AR Model)预测患者康复趋势;而后者,深入挖掘sEMG产生机理,通过单通道sEMG分解得到MUAPT,并采用特征提取、降维、聚类等方法,得到健康侧与患侧的类别中心距离,同样构建AR模型,预测患者康复趋势。本论文针对针刺治疗面瘫的20名患者开展临床实验,结果证明,两种方法都能得到较好的康复疗效量化评估及预测结果,准确率分别达到92.8%和90.94%以上。本文工作拓展了肌电信号分析的研究思路,对基于肌电信号的运动意图识别方法和康复疗效量化评估方法进行了深入探索,具有重要的理论价值和广阔的应用前景。
英文摘要: Surface electromyogram (sEMG) signal is a summation of motor unit action potential trains (MUAPTs), which can be readily measured on the skin during the muscle contraction. Moreover, it enables a non-invasive and objective quantification of neuromuscular function and motor control according to the recruitment and firing of the motor units (MUs). Hence, sEMG has been extensively used in human computer interface, prosthetic device control, disease diagnosis, pathologic analysis and rehabilitation evaluation. On the other hand, humans have gradually entered the aging society. Elderly people with limited mobility increase year by year due to the arthritis, stroke, Parkinson’s disease and so on. Moreover, over 50 billion persons per year suffer from physical disability as a result of earthquakes, car accidents etc. Consequently, more and more patients would like to receive physical rehabilitation to recover the mobility and limb’s function. Therefore, the incorporation of robot technology and rehabilitation medicine captures more and more researchers’ attention. Rehabilition robot advanced greatly during the past decades, but there are still many problems: 1) the existing rehabilitation robots are designed according to the principal of mechtronics and cannot recognize the patients’ movement intentions; 2) a few rehabilitation robots which can be controlled with physiological signals need multiple sensors and the signal recognition model should be trained respectively for each user. It reduces the efficiency of rehabilitation extremely; 3) the current rehabilitation assessment methods normally depend on the clinical experience of clinicians and the subjective feel of patients, which stifles the patients’ initiative to participate in rehabilitation. In this paper, we investigated the human movement intention recognition method and sEMG based rehabilitation quantitative assessment method. The concrete contents are as follows: Firstly, classification of gestures based on multi-channel sEMG has been investigated extensively. However, due to the nature of sEMG sensors, the more sensors are used, the greater chance for the sEMG to be influenced by environment noise. Furthermore, it is not feasible to use multi-sensors in some cases because of the bulky size of the sensors and the limited area of muscles. This paper proposes a novel sEMG recognition method based on the decomposition of single-channel sEMG. At first, sEMG is acquired when the participants do five predetermined hand gestures. Then, this signal is resolved into its component motor unit potential trains (MUAPTs), which includes 3 steps: weighted low pass differential (WLPD) filtering, spikes detection and clustering with Gaussian Mixture Model (GMM). After that, five MUAPTs are obtained and used for hand gestures classification, which includes 3 steps too: five features, integral of absolute value (IAV), maximum value (MAX), median of non-zero value (NonZeroMed) and semi-window energy (SemiEny1 and SemiEny2) are extracted to form feature matrix; this feature matrix is dimension reduced by non-negative matrix factorization (NMF); and the dimension-reduced feature matrix is classified with the algorithm of Linear Discriminate Analysis (LDA). Extensive experiments have been carried out to verify the performance of the propose method. The classification results indicate the proposed method can achieve an accuracy of 86.3% while the accuracy of traditional classification methods for single-channel sEMG is about 76%.Secondly, due to the different muscle shape, maximal voluntary contraction and fat content of different users, the morphology of sEMG varies greatly when different users do the same actions. Hence, the sEMG recognition model, trained with the sEMG data acquired from specific user, is merely applicable to the user himself; and for a novel user, the recognition model would be constructed again with his/her own data according to the aforementioned steps, which hinders the practical application of myoelectric interfaces immensely. In this paper, a sEMG recognition method which is applicable to multi-users is proposed. Firstly, single channel sEMG is decomposed into 30 MUAPTs, which includes four steps: two-order differential filter, threshold calculation, spike detection and hierarchical clustering. Secondly, the MUAPTs are updated with the templates orthogonalization; and Deep Belief Network is employed to classify the MUAPTs into five classes corresponding to the predefined five gestures. Six participants participated in this experiment to validate the effectiveness of the proposed method. Results indicated that this method can achieve a mean accuracy of 81.5%.Thirdly, traditional linear and statistical analysis methods have some significant limitations due to the short-term stationary and lower signal-noise ratio of sEMG. In this paper, we introduce chaotic analysis into the field of sEMG process to investigate the hidden nonlinear characteristics of sEMG of patients with facial paralysis. sEMG on the bilateral masseter, levator labii superiors and frontals of the 21 patients is recorded. Chaotic analysis is employed to extract new features, including correlation dimension, Lyapunov exponent, approximate entropy and so on. We discover the maximum Lyapunov exponents are all greater than 0, indicating that sEMG is a chaotic signal; correlation dimensions of sEMG on healthy sides are all smaller than that of diseased sides; and inversely, the approximate entropies of healthy sides are all greater than that of diseased sides. Chaotic analysis can provide a new insight into the complexity of the EMG and may be a vital indicator of diagnosis and recovery assessment of facial paralysis. Lastly, the current rehabilitation ef?cacy assessment is mainly based on empirical knowledge rather than scientific evidence, which stifles the patients’ initiative to participate in rehabilitation immensely. In this paper, two methods are proposed to quantitatively assess the effectiveness of rehabilitation based on sEMG features and sEMG decomposition respectively. For the method based on sEMG features, features in time and frequency domains are extracted and then dimension-reduced with PCA method; k-means algorithm is used to cluster the features into healthy and diseased subsets; and finally the center-to-center distance between the healthy and diseased sides are used to predict the recovery trend with an auto-regression model. For the sEMG decomposition method, single channel-sEMG is decomposed into several MUAPTs with hierarchical clustering method; and then, features such as IAV, MAX, NonZeroMed and SemiEny are extracted and dimension reduced with PCA method; finally, k-means algorithm and auto-regression model are also used to predict the recovery trend. Extensive experiment is carried out with 20 patients with facial paralysis to verify the performance of the proposed methods. Results indicate the proposed method can achieve an accuracy of 92.8% and 90.94% respectively in prediction of the recovery trend. This paper proposes novel sEMG analysis methods, especially for sEMG based movement intention recognition methods and treatment efficacy quantitative assessment methods. It has great theoretical significance and promising applications.
语种: 中文
产权排序: 1
内容类型: 学位论文
URI标识: http://ir.sia.cn/handle/173321/16790
Appears in Collections:机器人学研究室_学位论文

Files in This Item:
File Name/ File Size Content Type Version Access License
肌电信号分析方法研究及其在康复领域的应用.pdf(4001KB)----限制开放 联系获取全文
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[熊安斌]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[熊安斌]‘s Articles
Related Copyright Policies
Null
Social Bookmarking
Add to CiteULike Add to Connotea Add to Del.icio.us Add to Digg Add to Reddit
所有评论 (0)
暂无评论
 
评注功能仅针对注册用户开放,请您登录
您对该条目有什么异议,请填写以下表单,管理员会尽快联系您。
内 容:
Email:  *
单位:
验证码:   刷新
您在IR的使用过程中有什么好的想法或者建议可以反馈给我们。
标 题:
 *
内 容:
Email:  *
验证码:   刷新

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

 

 

Valid XHTML 1.0!
Copyright © 2007-2016  中国科学院沈阳自动化研究所 - Feedback
Powered by CSpace