表面肌电模式识别的新控制策略研究

发布时间：2017-12-26 22:29

本文关键词：表面肌电模式识别的新控制策略研究　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：人体的所有动作都是通过多块肌肉在神经控制系统的支配下互相协调共同完成的。伴随肌肉运动产生的肌电信号(Electromyography,EMG)是一种重要的电生理信号,可以实时地反映出肢体的运动意识和运动状态。依据这样的生理学基础,从放置于皮肤表面的电极捕获表面肌电信号(Surface EMG，，sEMG)实现的肌电控制技术一直是广受关注的研究课题,其在假肢控制、人机交互以及神经肌肉疾病康复治疗等领域都有着广泛的应用。从实现的控制策略角度,肌电控制已经从最初仅支持单自由度的开关控制、经比例控制发展到当前可实现复杂运动的多自由度控制,取得了许多令人瞩目的成就。特别是近年来兴起的基于模式识别的肌电控制策略,由于其革命性地扩展了肌电控制的可控自由度数量,已经成为了当前肌电控制领域的研究热点。本论文在表面肌电模式识别的基础上进行了深入探索,以人体最具灵巧性的手势动作为研究对象,提出了更贴近人体自然运动控制方式的新型肌电控制策略与相应实现方法,并通过实验验证了所提策略与方法的有效性。本论文的研究成果将有助于实现更加自然和谐的肌电控制系统,对肌电假肢控制、人机交互甚至脑卒中康复训练等研究与应用均具有重要指导意义。本文具体的研究工作与创新点主要有:(1)变化收缩力对肌电模式识别影响研究。此研究基于多通道电极阵列采集的高密度sEMG,系统地研究了不同的肌肉收缩力水平与肌电模式之间的关系,结合电极阵列所提供的空间特征,提出了适用于收缩力变化情况下,具有良好鲁棒性的肌电模式识别策略,并进一步设计了面向实用化的肌电控制方案。不同收缩力下的模式识别实验结果证实了所提出的肌电控制策略的有效性,本研究成果可以为实现适用于变力情况的肌电控制系统提供训练指导与电极安放的参考依据。(2)基于任务间共有协同的肌电模式识别研究。此研究的目的是探索动作任务间的共有协同在肌电模式识别中的应用。目前通过肌肉协同理论来解释人体运动的控制机制已经得到了越来越多研究人员的认可,而非负矩阵分解(NMF)在许多协同分析的研究中都有成功应用。本研究针对肌肉协同在肌电模式识别中的应用,提出了任务间共有协同提取的判别非负矩阵分解(DNMF)算法。研究对12类手势动作进行了分类识别,并与传统的协同提取算法作比较,实验结果证实了判别非负矩阵分解在肌电模式识别中的可行性与有效性。(3)基于任务特有协同的多自由度并行控制策略研究。此研究的目标是实现精细手指运动的多自由度并行控制。研究以灵巧的五指运动为研究对象,每个手指对应一个自由度,提出了多自由度的并行控制策略。在单一自由度水平上,针对手指的不同状态利用NMF提取特有协同,依据特有协同重构样本特征的失真程度来判定该自由度运动状态。最终,通过对多个自由度运动状态决策的融合判定手指运动模式。为了验证所提方法的有效性,对健康受试者多手指组合运动的多类手势动作进行了多自由度的并行识别,并与传统的模式识别算法(串行控制策略),以及所提并行控制框架集成常规分类器的方法做了比较分析。实验结果证实了基于任务特有协同的多自由度并行控制策略不仅更接近人体运动的自然控制方式,而且具有更高的控制性能。进一步的,研究还通过脑卒中患者的肌电数据验证了所提并行控制策略在偏瘫手臂运动识别中的有效性,展示了所提并行控制策略在脑卒中康复训练领域的应用前景。
[Abstract]:All the movements of the human body are accomplished by the coordination of multiple muscles under the control of the neural control system. Electromyography (EMG) is an important electrophysiological signal, which can reflect the movement consciousness and motion state of limbs in real time. On the basis of studying such physiological capture surface EMG signals from electrodes placed on the surface of the skin (Surface EMG sEMG) technology to realize the myoelectric control has been a research topic of wide concern, the prosthetic control and human-computer interaction and rehabilitation treatment of neuromuscular disease and other fields have a wide range of applications. From the perspective of control strategy, EMG control has developed from the initial control of single degree of freedom switch control and proportional control to the multi degree of freedom control of complex motion, which has made many notable achievements. In recent years, especially the EMG control strategy based on pattern recognition, because it revolutionized the number of controllable degrees of freedom of EMG control, has become a research hotspot in the field of EMG control. This thesis makes a deep exploration based on surface EMG pattern recognition, gesture to most human dexterity as the research object, put forward the model closer to the natural body movement EMG control mode control strategy and corresponding implementation method, and the effectiveness of the proposed strategy and method is verified by experiment. The research results in this paper will help to achieve more natural and harmonious EMG control system, which is of great guiding significance for the research and application of EMG prosthesis control, human-machine interaction, and stroke rehabilitation training. The main research work and innovation of this paper are as follows: (1) the influence of the change contractile force on the pattern recognition of EMG. This study is based on the high density sEMG multichannel electrode array acquisition, systematically studies the relationship between muscle force and EMG pattern of different spatial features with electrode array is provided, put forward for the contraction force change, EMG pattern recognition method has good robustness, and further design the EMG for the practical control scheme. The results of pattern recognition experiments with different contractions confirm the effectiveness of the proposed EMG control strategy. The research results can provide references for training and placement of EMG control systems for variable force situations. (2) the study of EMG pattern recognition based on co - task cooperation among tasks. The purpose of this study is to explore the application of common synergy between action tasks in EMG pattern recognition. At present, the control mechanism of human motion has been recognized by more and more researchers through muscle synergy theory. Instead of negative matrix factorization (NMF), it has been successfully applied in many collaborative analysis researches. This research aims at the application of muscle synergy in EMG pattern recognition, and proposes a discriminant non negative matrix decomposition (DNMF) algorithm. 12 kinds of gesture actions are classified and identified, and compared with the traditional collaborative extraction algorithm. The experimental results confirm the feasibility and effectiveness of discriminant non negative matrix factorization in EMG pattern recognition. (3) research on multi degree of freedom parallel control strategy based on task specific synergy. The aim of this study is to realize the multi degree of freedom parallel control of fine finger motion. The study takes the dexterous five fingered motion as the research object. Each finger corresponds to a degree of freedom, and a parallel control strategy of multiple degrees of freedom is proposed. At a single degree of freedom level, for the different states of fingers, we use NMF to extract specific synergy, and determine the motion state of the degree of freedom based on the degree of distortion of specific cooperative reconstruction. Finally, the motion pattern of the finger is determined by the fusion of multiple freedom motion state decisions. In order to verify the validity of the method, many kinds of gestures of healthy subjects were combined multi finger movement parallel identification of multiple degrees of freedom, and the traditional pattern recognition algorithm (serial control strategy), and the proposed method of concurrency control framework integrated conventional classifier to do a comparative analysis. The experimental results confirm that the multi degree of freedom parallel control strategy based on task specific cooperation is not only closer to the natural control mode of human motion, but also has higher control performance. Further, the validity of the proposed parallel control strategy in the recognition of hemiplegic arm motion is verified by EMG data of stroke patients, showing the application prospect of the parallel control strategy in the field of stroke rehabilitation training.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.7;R318

【相似文献】