根据肌电图（EMG）信号评估人手臂肌肉力和肌肉疲劳的人工智能方法

发布时间：2018-03-10 16:00

本文选题：表面肌电信号　切入点：肌力　出处：《华中科技大学》2013年博士论文　论文类型：学位论文

【摘要】：由神经系统和肌骨骼系统所组成的人体运动系统使人们可以完成各种复杂的任务。其中,骨骼肌是驱动人体运动的力量源泉,它接收大脑发送的并经由脊髓神经传输的电信号命令,产生相应的收缩并带动关节运动。按照一定时空顺序排列的电信号命令可以驱使多肌肉群组协调地收缩,从而产生协调的人体运动。肌骨骼肌系统的任何缺陷均会影响人体运动和执行任务的能力。鉴于骨骼肌在人体运动系统中的重要地位,骨骼肌及其机能的研究逐渐引起人们的研究兴趣。肌力和肌肉疲劳是肌肉机能评价的两项重要指标,其评估结果有助于认识人们的运动能力。但如何准确地评估人体运动过程中的肌力和肌肉疲劳仍是一个有待解决的问题。肌电信号是一种记录肌肉收缩电信号的技术,也是一种评估肌肉活动的有效方法。然而,在处理肌电信号来评估肌力和肌肉疲劳的过程中,也还存在许多问题和难题。该工作旨在解决以下三个问题：第一,对肌力的评估过程往往包含两个步骤：(1)从肌电信号中提取肌肉活动特性,(2)采用Hill模型计算肌力的大小。这一过程除了要优化许多模型参数之外,计算过程本身也很复杂。在这项工作中,我们提出了两种新的肌力评估模型,简化了评估肌力的方法。第一种模型采用神经网络理论,第二种模型则以模糊逻辑为基础。我们开展了人体手臂六组肌肉群组肌电采集实验,并利用MATLAB对所提出的两种模型进行了仿真研究。我们对所提出的两种模型进行了计算速度和稳定性的比较。研究表明,模糊逻辑模型在计算速度和稳定性方面比神经网络模型更快、更稳定,但是其结构比神经网络模型更为复杂。第二,肌电信号的记录和处理是研究肌肉活动的基本步骤,因为肌电原始信号往往会受到大量噪声的影响,为提取有效、准确的肌电活动信号,人们往往采用各种信号预处理技术进行滤波、归一化等处理。虽然这一过程易于实现,但该过程可能降低信号数据的准确性。为了避免信号处理技术的这一影响,我们提出了一种遗传算法技术,将原始肌电信号作为模型的输入,而模型的输出则是一种能用于评估肌肉力量和肌肉疲劳的信号。为了验证这种模型的有效性,我们首先采用了带遗传算法的神经网络模型评估肌力水平,然后与原神经网络模型肌力评估方法进行比较,结果显示,两种模型对肌力评估的效果很相似并且都具有收敛性。最后,肌肉的疲劳会直接导致肌力的下降,是影响肌肉活动能力的一个重要因素。我们提出了一种根据模糊逻辑理论评估肌肉疲劳的新方法。这种方法不仅有助于肌肉疲劳的评估,而且适用于任何手臂肌力的评估。这种模型同样由原始肌电信号驱动,无需采用任何信号预处理。与前面肌力评估模型不同的是,这种模型采用的是模糊逻辑方法,而不是遗传算法。通过对五十位志愿者上臂主动肌/拮抗肌对(肱二头肌和肱三头肌)肌肉疲劳的实验验证,证明了这种方法用于评估肌肉疲劳的有效性。此外,我们还研究了肌肉疲劳与受试者年龄之间的关系,直观地反映了肌肉抗疲劳特性随年龄增长而降低的发展趋势。
[Abstract]:Human motion system is composed of the nervous system and musculoskeletal system so that people can accomplish complex tasks. Among them, the skeletal muscle is the source of power to drive the human motion, it receives brain transmission and transmission of electrical signals through the spinal cord to produce the corresponding command, shrink and promote joint movement. According to a certain order of time and space the signal command can drive much muscle contraction to produce group coordination, coordination of human motion. The ability of any defect of muscle system will affect the human movement and mission. In view of the important status of skeletal muscle in the human body movement in the system of skeletal muscle and its function has gradually aroused people's interest in the study. Muscle strength and muscle fatigue are two important indicators of muscle function evaluation, the evaluation result is helpful to understand the movement of people's ability. But how to accurately assess the human. Muscle strength and muscle fatigue in the process of movement are still a problem to be solved.
The EMG signal is a record of muscle contraction signal technology, is a kind of effective method to evaluate the muscle activity. However, in the treatment of EMG signal to evaluate the process of muscle strength and muscle fatigue, there are many problems and difficulties. This work aims to solve the following three questions:
First, the evaluation process of muscle strength often consists of two steps: (1) extraction of muscle activity characteristics from EMG, muscle strength (2) calculated using the Hill model. The size of this process in addition to the optimization of many parameters in the model, the calculation process is very complicated. In this work, we propose two models for evaluation of muscle strength of new, simplified method of evaluating the muscle strength. The first model uses the neural network theory, second kinds of model based on fuzzy logic. We carried out six groups of human arm muscle EMG acquisition experiment group, and two of the proposed model is simulated by MATLAB. We were comparison of calculation speed and stability of two kinds of the proposed model. The results show that the fuzzy logic model in computing speed and stability than the neural network model is faster, more stable, but its structure than the neural network model It's more complicated.
Second, recording and processing of the EMG signal is a basic step of muscle activity, because the original EMG signals often affected by noise, extracting effective EMG signals accurately, people often use a variety of signal preprocessing filtering, normalization processing. Although this process is easy to implement, but the process may reduce the accuracy of signal data. In order to avoid the influence of signal processing technology, we propose a genetic algorithm, the original EMG signal as the input of the model, and the output of the model is a kind of signal can be used to assess muscle strength and muscle fatigue. In order to verify the validity of this model, we first by using the neural network model with genetic algorithm to evaluate strength level, and then compare with the original neural network model strength evaluation method showed that two kinds of model The effectiveness of the muscle strength assessment is similar and convergent.
Finally, muscle fatigue will directly lead to the decline of muscle strength, is an important factor affecting muscle activity. We propose a new method based on the fuzzy logic theory to evaluate muscle fatigue. This method not only helps to evaluate the muscle fatigue assessment, but also suitable for any arm muscle strength. This model is driven by the same the original EMG signal, without using any signal processing. Unlike the previous strength assessment model, this model is based on fuzzy logic method, rather than genetic algorithm. Based on fifty volunteers active arm muscle / antagonist (biceps and triceps) on experimental verification of muscle fatigue, the method for evaluating the effective muscle fatigue is proved. In addition, we also studied the relationship between muscle fatigue and the age of the subjects, can reflect the muscle fatigue resistance increased with aging. A long and low trend of development.

【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：R741.044;TP18

【相似文献】