基于sEMG信号的上肢康复机器人训练系统的研究

发布时间：2018-05-05 15:11

本文选题：sEMG信号 + 上肢康复机器人　；参考：《天津理工大学》2017年硕士论文

【摘要】：表面肌电信号(surface electromyography,sEMG)是通过表面电极在人体皮肤表面采集的肌肉运动过程中所产生的电位信号。人体在执行不同动作过程中作用的肌肉群是不同的,因此产生的电信号也是不同的。本文通过对人体手臂表面肌电信号的处理来识别人体上肢的动作模式,从而针对偏瘫患者实现利用偏瘫患者健测上肢引导康复机器人辅助偏瘫患者患侧上肢执行康复运动训练。本研究致力于研究一套基于sEMG信号的上肢康复机器人康复训练及评价系统,系统主要包括表面肌电信号采集设备、上肢外骨骼机械臂、计算机、电机及控制单元等设备。本研究针对偏瘫患者,提出了一种基于sEMG信号自主式的上肢康复训练系统来帮助患者实现主动训练。研究将模式识别技术和机器人技术相结合,提出了整个系统的设计方案:通过采集偏瘫患者健测上肢执行六类不同动作模式下的sEMG信号,首先对sEMG信号进行预处理,然后对s EMG信号进行特征提取以及模式识别分类,针对参与实验的八名受试者得出离线模式识别平均分类准确率91%。上肢自主在线康复训练中通过偏瘫患者健侧执行的某一动作模式根据模式识别离线训练好的分类器得出分类结果驱动电机控制康复机器人执行相应的动作,实现自主康复训练。八位受试者动作执行的最高准确率高达98.3%,平均准确率达到92%。本研究还提出了一种以肌力大小为标准的、一种定量的、客观的康复评价方法,用于对康复训练过程中患者肌力大小的恢复程度做出客观评价。通过采集不同肌力大小的的表面肌电信号,并按临床医学肌力分级表将肌力分为6个等级。由于肌肉电信号的非线性特性,不同肌力水平下的肌电信号的复杂度不同,本研究选用模糊近似熵对其进行建模分析,并选取肌电信号的模糊近似熵和肌电强度组成的二维特征向量采用有向无环图支持向量机分类方法进行分类,分类正确率达到92%。本研究对于上肢康复医学领域康复训练方式和评价方法提供了参考,有助于拓宽相关领域的研究思路。
[Abstract]:Surface electromyography (EMG) is a potential signal generated during muscle movement on the surface of human skin. The muscles that the human body acts on in different actions are different, so the electrical signals are different. In this paper, we recognize the movement pattern of human upper limb by processing the EMG signal on the surface of human arm, so as to realize the rehabilitation exercise training of the affected upper limb of the hemiplegic patient by using the rehabilitation robot to guide the upper limb of the hemiplegic patient by using the healthy test of the upper limb of the patient with hemiplegia. This study is devoted to the study of an upper limb rehabilitation robot rehabilitation training and evaluation system based on sEMG signal. The system mainly includes surface electromyographic signal acquisition equipment, upper limb exoskeleton manipulator, computer, motor and control unit and so on. In this study, a sEMG signal independent upper limb rehabilitation training system was proposed to help patients with hemiplegia to achieve active training. The design scheme of the whole system is put forward by combining the pattern recognition technology with the robot technology. Firstly, the sEMG signal is preprocessed by collecting the sEMG signals of the upper limbs of hemiplegic patients and performing six kinds of different action modes. Then the feature extraction and pattern recognition classification of s EMG signal were carried out, and the average classification accuracy of offline pattern recognition was obtained for the eight subjects involved in the experiment. In the automatic and online rehabilitation training of upper limb, a certain action pattern performed by the healthy side of the hemiplegic patient was obtained by the classifier of off-line training. The classification result was obtained by driving the motor to control the rehabilitation robot to perform the corresponding action, and to realize the autonomous rehabilitation training. The highest accuracy rate was 98.3%, with an average accuracy rate of 92%. In this study, a quantitative and objective rehabilitation evaluation method based on muscle strength is proposed, which can be used to evaluate the recovery degree of muscle strength of patients during rehabilitation training. The surface EMG signals of different muscle strength were collected and divided into 6 grades according to the clinical strength scale. Because of the nonlinear characteristics of EMG signals, the complexity of EMG signals under different muscle force levels is different. In this study, fuzzy approximate entropy is used to model and analyze EMG signals. The two dimensional eigenvector which is composed of fuzzy approximate entropy and myoelectric intensity of EMG signal is classified by directed acyclic graph support vector machine. The classification accuracy is 92%. This study provides a reference for rehabilitation training methods and evaluation methods in the field of upper limb rehabilitation medicine, and helps to broaden the research ideas in related fields.
【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242

【相似文献】