表面肌电信号分解算法及其在小儿脑瘫评诂中的应用

发布时间：2018-05-23 17:01

本文选题：表面肌电信号 + 表面肌信号模型　；参考：《中国科学技术大学》2012年博士论文

【摘要】：表面肌电信号(sEMG)是由肌肉兴奋时所募集的运动单位产生的一个个动作电位序列(Motor Unit Action Potential trains, MUAPT)在皮肤表面叠加而成,是一种非平稳的微弱信号。而肌电信号的分解则是其形成的求逆过程,从分解结果可以获得运动单位的募集发放信息和运动单位动作电位(MUAP)的波形信息,这是提供观察中枢神经系统各运动神经元活动状态的一种途径,由此可以获得神经肌肉疾病患者运动控制机制病理等变化的正确理解和解释。表面肌电信号分解是一项难度较大的研究工作,研究成果对于神经肌肉系统控制的基础研究和神经肌肉疾病的临床诊断具有重要的理论意义和应用价值。表面肌电信号是一种复杂的电信号,对其进行分解是一项难度很大的工作。本文首先进行了表面肌电信号的仿真研究,并在针对插入式肌电信号与表面肌电信号进行广泛调研之后,提出了两种实现表面肌电信号分解的新方法,同时通过对仿真信号和真实信号的实验研究,论证算法的可行性与有效性。另外,利用表面肌电信号对痉挛型脑瘫患儿的运动神经元发放特性进行了研究,旨在为脑瘫患儿的运动障碍评估提供一种量化指标。本文的主要工作和研究成果有： 1. sEMG信号的仿真研究。根据表面肌电信号的生理层模型,实现了对表面肌电信号的仿真,用以提供验证肌电信号分解效果的仿真信号。重点讨论了检测电极与肌纤维的相对位置以及肌纤维的传导速度对单纤维动作电位产生的影响,同时提出了一种动态恒力的运动单元募集方案。sEMG信号的仿真根据真实采集方案进行了单电极和差分式电极两种电极方案的仿真,在一定程度上描述了sEMG信号的形成过程。 2.针对盲源分离求解欠定混合方程存在的问题,提出将基于稀疏分量分析的欠定盲源分离算法用于肌电信号分解的实现方法。在去噪和稀疏化表面肌电信号的过程中,采用了一种基于遗传算法改进的匹配追踪算法,提高了稀疏化和去噪的效率；采用hough变换法进行混合矩阵估计,并提出一种空间退化方法优化了hough算法,减少了伪轴的产生；在聚类算法中,采用减法聚类实现了对混合矩阵的自动估计。实验结果表明,基于稀疏分量分析的欠定盲源分离方法可以较好的解决低收缩力下表面肌电信号的分解问题。 3.提出了基于多维子空间聚类的独立分量分析实现表面肌电信号分解的算法。针对表面肌电信号在中度收缩力下难以估计源信号的难点,通过将多维的表面肌电信号投影在多维子空间中进行降维,获得中度收缩力下的表面肌电信号的局部混合结构,从而估计混合矩阵,并由此获得表面肌电信号源信号的估计值。同时进行仿真信号与真实信号的验证实验。在真实表面肌电信号的实验中,采用双源法进行对比实验。实验结果表明,本文算法对于中度收缩力的表面肌电信号分解具有明显的分解效果。 4.提出利用肌电信号运动单位动作电位数目估计的方法研究痉挛型脑瘫患儿神经元的发放特性。小儿脑瘫患者的运动障碍分级和康复评估具有重要的临床价值。本文采用平滑非线性能量算子(SNEO)算法,对痉挛型脑瘫患儿表面肌电信号中的运动单位动作电位(MUAP)数目进行估计,并获得MUAP的平均发放间隔(IPI)。对十四名各不同运动障碍级别脑瘫患者的实验结果显示,其肌电信号MUAP的IPI与他们活动度的级别即运动障碍的程度呈正相关的关系,且具有明显的差异。实验结果表明本文方法的有效性,采用IPI参数能够反映脑瘫患儿的运动障碍程度。本论文研究得到国家自然科学基金项目“基于线性变化力采集和多通道时空信息的表面肌电信号分解”(30870656)的资助。
[Abstract]:Surface EMG signal ( sEMG ) is a kind of non - stationary weak signal generated by motor unit action potential ( MUAPT ) generated by muscle excitation . The decomposition of myoelectric signal is a non - stationary weak signal . It is a way to observe the activity of motor neuron in the central nervous system . It is a very difficult research work . The research results have important theoretical significance and application value for the basic research of neuromuscular system control and the clinical diagnosis of neuromuscular disease .

The surface myoelectric signal is a complex electrical signal . It is a difficult task to decompose the surface EMG signal . After extensive investigation of the EMG signal and the surface myoelectric signal , two new methods for realizing the surface myoelectric signal decomposition are presented . At the same time , the feasibility and effectiveness of the algorithm are demonstrated by the experimental research on the simulated signal and the real signal .

The main work and research results of this paper are as follows :

1 . The simulation of EMG signal is studied . Based on the physiological layer model of surface myoelectric signal , the simulation of surface EMG signal is realized . The influence of the relative positions of the detection electrode and the muscle fiber and the conduction velocity of the muscle fiber on the action potential of the single fiber is discussed . The simulation of the two electrode schemes of single electrode and differential electrode is carried out according to the real acquisition scheme . The formation process of sEMG signal is described to a certain extent .

2 . Aiming at the problem of solving the problem of under - definite mixing equation based on sparse component analysis , this paper proposes a method to realize the decomposition of myoelectric signal based on sparse component analysis . In the process of denoising and thinning the surface myoelectric signal , a matching pursuit algorithm based on genetic algorithm is adopted to improve the efficiency of thinning and de - noising ;
The hough transform method is used to estimate the mixed matrix , and a space degradation method is proposed to optimize the hough algorithm , which reduces the generation of pseudo - axes ;
In the clustering algorithm , the automatic estimation of mixed matrix is realized by means of subtraction clustering . The experimental results show that the problem of decomposition of the surface myoelectric signal under the low shrinkage force can be better solved by using the sparse component analysis .

3 . An algorithm for the decomposition of surface myoelectric signal based on multi - dimensional subspace clustering is presented . It is difficult to estimate the source signal under moderate contraction force for surface myoelectric signal . By using the multi - dimensional surface myoelectric signal projection in the multi - dimensional subspace , the local hybrid structure of the surface EMG signal is obtained . The simulation signal and the real signal verification experiment are carried out . The experiment results show that the algorithm has obvious decomposition effect on the surface myoelectric signal decomposition of moderate contraction force .

4 . The method of estimating the number of motor potential ( MUAP ) in children with cerebral palsy was studied by using the method of estimating the number of motor potentials in children with cerebral palsy . The results showed that the IPI of MUAP was positively correlated with the level of activity and the degree of motion disturbance . The results showed that the effectiveness of the method was effective , and the IPI parameters could reflect the degree of motor impairment in children with cerebral palsy .

This thesis is funded by the National Natural Science Foundation Project " Decomposition of Surface EMG Signal Based on Linear Variation Force Acquisition and Multi - channel Temporal and Temporal Information " ( 30870 656 ) .
【学位授予单位】：中国科学技术大学
【学位级别】：博士
【学位授予年份】：2012
【分类号】：R742.3

【相似文献】