动作表面肌电信号的非线性特性研究

发布时间：2018-05-10 07:10

本文选题：表面肌电信号 + 非线性分析　；参考：《上海交通大学》2012年硕士论文

【摘要】：人体动作电信号由神经和肌肉组成的运动单元产生,然后控制肌肉进行协同作用,从而完成人体动作。不同的电信号驱动不同的动作。在动作完成的整个过程中,这些电信号通过人体组织在皮肤表面上输出,被皮肤处的电极等设备采集到。采集到的电信号称为表面肌电信号。表面肌电信号与肌肉活动情况和功能特性之间存在着不同程度的关联性,在一定程度上反映了神经肌肉的状况和活动情况。因此,表面肌电信号在临床医学、运动医学、人机工效学、康复医学、神经生理学、电生理学等领域被广泛应用。目前,关于动作表面肌电信号的非线性特性研究还处于初级探索阶段。在已有的表面肌电信号采集方法和技术的基础上,本文了设计了表面肌电信号的采集实验,采集到人体前臂内翻、外翻、握拳、展拳、上切和下切六类动作表面肌电信号作为研究对象。利用非线性时间序列分析方法对表面肌电信号的非线性特性进行研究,验证了表面肌电信号的非线性特性,表明动作表面肌电信号是一混沌信号。这对于深入认识神经肌肉系统的功能活动规律及其实质、建立更加科学合理的肌肉功能非损伤性评价技术均具有重要的价值。为了进一步了解动作表面肌电信号的非线性特性,本文主要利用小波变换和希尔伯特-黄变换对动作表面肌电信号进行了多尺度分解,进而对每一个尺度上的动作表面肌电信号的非线性特性进行了研究,把动作表面肌电信号的非线性特性在不同的尺度上进行展开,更加了解了动作表面肌电信号的非线性特性。最后,为了提高动作表面肌电信号的识别率,本文提出一种将非线性分析和多尺度分析结合的方法。该方法从其非线性和非平稳特性的角度出发,引入了多尺度非线性特征,并应用到人体前臂六类动作表面肌电信号的模式识别中。将多尺度非线性特征输入支持向量机,并结合核主元分析方法,使动作表面肌电信号的平均识别率达到98%。结果表明,利用多尺度非线性特征对动作表面肌电信号进行模式识别效果良好。
[Abstract]:Human action signals are generated by motor units composed of nerves and muscles, and then control the muscles for synergistic action, thus accomplishing human actions. Different electrical signals drive different movements. In the whole process of action, these electrical signals are output from human tissues on the skin surface and collected by devices such as electrodes in the skin. The collected electrical signals are called surface EMG signals. There is some correlation between surface EMG signal and muscle activity and functional characteristics, which reflects the condition and activity of neuromuscular to some extent. Therefore, surface electromyography is widely used in clinical medicine, sports medicine, ergonomics, rehabilitation medicine, neurophysiology, electrophysiology and so on. At present, the study of nonlinear characteristics of action surface EMG signal is still in the primary stage. On the basis of existing methods and techniques of collecting surface EMG signals, this paper designs an experiment to collect surface EMG signals, which can collect human forearm varus, valgus, clenched fist, extended fist, and so on. Six types of action surface electromyography (EMG) were studied. The nonlinear characteristics of surface EMG signal are studied by using nonlinear time series analysis method, and the nonlinear characteristics of SEMG signal are verified, which indicates that the action surface EMG signal is a chaotic signal. This is of great value for further understanding the functional activity and essence of neuromuscular system and establishing a more scientific and reasonable non-injurious evaluation technique for muscle function. In order to further understand the nonlinear characteristics of action surface EMG signal, wavelet transform and Hilbert-Huang transform are mainly used to decompose the action surface EMG signal. Furthermore, the nonlinear characteristics of EMG signals on each scale are studied, and the nonlinear characteristics of EMG signals are expanded on different scales to understand the nonlinear characteristics of EMG signals. Finally, in order to improve the recognition rate of EMG signals, a method combining nonlinear analysis and multi-scale analysis is proposed. From the point of view of its nonlinear and non-stationary characteristics, this method introduces multi-scale nonlinear features, and is applied to the pattern recognition of six kinds of EMG signals on the forearm. The multi-scale nonlinear feature is input into the support vector machine and the kernel principal component analysis method is used to make the average recognition rate of the EMG signal on the action surface up to 98. The results show that the pattern recognition of EMG signals on the action surface is effective by using multi-scale nonlinear features.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH772

【共引文献】