基于表面肌电信号的人体下肢运动自动分类研究
发布时间：2018-09-05 10:30
【摘要】：表面肌电(surface Electromyography, sEMG)信号是一种复杂的人体表皮下肌肉电活动在皮肤表面处时间和空间上的综合结果,是从人体骨骼肌表面通过非侵入方式记录下来的神经肌肉活动时发放的生物电信号,它能在非损伤状态下实时反映神经肌肉的功能状态。本文主要研究人体下肢sEMG信号的采集与处理以及基于sEMG信号的运动模式辨识方法,研究内容主要涉及神经—肌肉学科中的神经肌电信号、信号处理和模式识别等方面。 随着材料、传感器和计算机等技术发展,国内外对表面肌电的研究也逐步深入,使得表面肌电信号不仅在运动医学、临床医学及康复医学等领域被广泛应用,而且还成为了人工假肢的理想控制信号。sEMG信号的模式识别是其应用的基础,为此,本文深入探讨了如何由采集的sEMG信号来识别下肢不同的运动模式,其目的就是基于sEMG信号的非平稳性和随机性,运用现代信号处理方法寻求其内在的本质特征,并深入研究及运用现代模式识别理论设计模式分类器,使其能够对下肢运动模式的本征值进行有效识别,为揭示sEMG信号的本质与多自由度肌电控制假肢的实用化提供理论依据,主要工作和创新之处如下： 1.设计了sEMG信号的放大滤波电路,sEMG信号的放大滤波电路是实现肌电信号采集系统的关键,根据sEMG信号的幅频特性和外界信号对其影响,本文设计了较好的滤波电路,特别是设计出了一种新型的50Hz工频陷波电路,能够很好地解决工频噪声对sEMG信号的不良影响； 2.利用小波变换的多分辨率分析技术,利用小波分解系数矩阵的奇异值构建sEMG信号特征向量,结合BP神经网络和支持向量机分类器,对人体下肢常见六种运动的分类进行了研究,并完成了对下肢肌肉的疲劳评估和基于sEMG信号的路况识别； 3.提出了基于粒子群优化的过程神经网络分类算法,兼顾了采集信号的空间耦合作用和时间积累效应,避免了因特征提取而导致的信息流失,在无需进行特征提取的情况下,完成了对人体下肢不同运动模式的自动分类。与传统的过程神经网络算法相比,本文采用的粒子群算法通过对网络系数的优化,大大提高了算法的执行效率,并取得了理想的分类准确率。
[Abstract]:Surface electromyography (surface Electromyography, sEMG) signal is a complex subepidermal muscle electrical activity on the surface of the skin time and space synthesis results. It is a bioelectric signal which is recorded from the surface of human skeletal muscle in a non-invasive manner when the neuromuscular activity is recorded. It can reflect the functional state of the neuromuscular in a non-invasive state in real time. This paper mainly studies the acquisition and processing of human lower limb sEMG signal and the method of motion pattern identification based on sEMG signal. The research content mainly involves neuromyogram signal, signal processing and pattern recognition in neuromuscular discipline. With the development of materials, sensors and computers, the research on surface electromyography at home and abroad has gradually deepened, which makes surface electromyography widely used not only in sports medicine, clinical medicine and rehabilitation medicine, but also in the fields of sports medicine, clinical medicine and rehabilitation medicine. Moreover, the pattern recognition of the ideal control signal. SEMG signal of artificial prosthesis is the basis of its application. For this reason, this paper deeply discusses how to recognize the different motion modes of lower extremity by the collected sEMG signal. Based on the non-stationary and randomness of sEMG signal, the purpose of this paper is to use modern signal processing method to find out its intrinsic characteristics, and to design pattern classifier based on modern pattern recognition theory. It can effectively identify the eigenvalues of lower extremity motion mode and provide theoretical basis for revealing the nature of sEMG signal and the practicality of multi-degree-of-freedom myoelectric control prosthesis. The main work and innovations are as follows: 1. The amplifying and filtering circuit of sEMG signal is the key to realize the EMG signal acquisition system. According to the amplitude and frequency characteristic of sEMG signal and the influence of external signal, a better filter circuit is designed in this paper. In particular, a new type of 50Hz power frequency notch circuit is designed, which can solve the adverse effects of power frequency noise on sEMG signal. 2. Using the multi-resolution analysis technology of wavelet transform, using the singular value of wavelet decomposition coefficient matrix to construct the characteristic vector of sEMG signal, combining with BP neural network and support vector machine classifier, the classification of six common movements of human lower limb is studied. The fatigue evaluation of lower extremity muscle and the recognition of road condition based on sEMG signal are completed. 3. A process neural network classification algorithm based on particle swarm optimization (PSO) is proposed, which takes into account the spatial coupling effect and time accumulation effect of collected signals, and avoids the loss of information caused by feature extraction. The automatic classification of different movement modes of human lower extremities has been completed. Compared with the traditional process neural network algorithm, the particle swarm optimization algorithm proposed in this paper greatly improves the efficiency of the algorithm by optimizing the network coefficients, and achieves the ideal classification accuracy.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R318.04;TN911.7
本文编号：2223999
[Abstract]:Surface electromyography (surface Electromyography, sEMG) signal is a complex subepidermal muscle electrical activity on the surface of the skin time and space synthesis results. It is a bioelectric signal which is recorded from the surface of human skeletal muscle in a non-invasive manner when the neuromuscular activity is recorded. It can reflect the functional state of the neuromuscular in a non-invasive state in real time. This paper mainly studies the acquisition and processing of human lower limb sEMG signal and the method of motion pattern identification based on sEMG signal. The research content mainly involves neuromyogram signal, signal processing and pattern recognition in neuromuscular discipline. With the development of materials, sensors and computers, the research on surface electromyography at home and abroad has gradually deepened, which makes surface electromyography widely used not only in sports medicine, clinical medicine and rehabilitation medicine, but also in the fields of sports medicine, clinical medicine and rehabilitation medicine. Moreover, the pattern recognition of the ideal control signal. SEMG signal of artificial prosthesis is the basis of its application. For this reason, this paper deeply discusses how to recognize the different motion modes of lower extremity by the collected sEMG signal. Based on the non-stationary and randomness of sEMG signal, the purpose of this paper is to use modern signal processing method to find out its intrinsic characteristics, and to design pattern classifier based on modern pattern recognition theory. It can effectively identify the eigenvalues of lower extremity motion mode and provide theoretical basis for revealing the nature of sEMG signal and the practicality of multi-degree-of-freedom myoelectric control prosthesis. The main work and innovations are as follows: 1. The amplifying and filtering circuit of sEMG signal is the key to realize the EMG signal acquisition system. According to the amplitude and frequency characteristic of sEMG signal and the influence of external signal, a better filter circuit is designed in this paper. In particular, a new type of 50Hz power frequency notch circuit is designed, which can solve the adverse effects of power frequency noise on sEMG signal. 2. Using the multi-resolution analysis technology of wavelet transform, using the singular value of wavelet decomposition coefficient matrix to construct the characteristic vector of sEMG signal, combining with BP neural network and support vector machine classifier, the classification of six common movements of human lower limb is studied. The fatigue evaluation of lower extremity muscle and the recognition of road condition based on sEMG signal are completed. 3. A process neural network classification algorithm based on particle swarm optimization (PSO) is proposed, which takes into account the spatial coupling effect and time accumulation effect of collected signals, and avoids the loss of information caused by feature extraction. The automatic classification of different movement modes of human lower extremities has been completed. Compared with the traditional process neural network algorithm, the particle swarm optimization algorithm proposed in this paper greatly improves the efficiency of the algorithm by optimizing the network coefficients, and achieves the ideal classification accuracy.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R318.04;TN911.7
【参考文献】
相关期刊论文 前10条
1 贾雪琴;王旭;李景宏;杨丹;;基于小波变换和K-L展开的单通道表面肌电信号识别[J];东北大学学报;2006年08期
2 王宏;姬彦巧;赵长宽;李琪;;基于肌肉电信号控制的假肢用机械手的设计[J];东北大学学报;2006年09期
3 杜春梅,田丰,崔建国;用小波和神经网络相结合的方法识别人体表面肌电信号[J];沈阳航空工业学院学报;2005年03期
4 王飞,罗志增;基于AR模型和BP网络的表面EMG信号模式分类[J];华中科技大学学报(自然科学版);2004年S1期
5 姬彦巧;王宏;赵长宽;李封;;假肢用机械手的机构设计与运动学分析[J];机械设计;2006年06期
6 胡天培;肌电特征发现与肌电康复研究[J];上海交通大学学报;1994年03期
7 吴冬梅;孙欣;张志成;杜志江;;表面肌电信号的分析和特征提取[J];中国组织工程研究与临床康复;2010年43期
8 高关心,夏慧琳;生物医学工程学与最新发展技术[J];医疗设备信息;2001年08期
9 刘志泉;我国肢体残疾人概况[J];中国康复医学杂志;2003年08期
10 贾晓枫,陈统一,陈中伟,张键,张晓文,斯扬,胡天培,高忠华,杨煜普;中国首例人体残肢神经信息控制电子假肢实验研究快报[J];中华物理医学与康复杂志;2004年01期
，本文编号：2223999
本文链接：https://www.wllwen.com/yixuelunwen/swyx/2223999.html
