基于脑电和肌电的假手多自由度动作识别和控制方法研究

项目名称： 基于脑电和肌电的假手多自由度动作识别和控制方法研究

项目编号： No.60874102

项目类型： 面上项目

立项/批准年度： 2009

项目学科： 生物科学

项目作者： 罗志增

作者单位： 杭州电子科技大学

项目金额： 33万元

中文摘要： 丧失肢体的残疾人渴望拥有像正常人一样能按自身意愿支配的假肢，所以，高性能电动假肢研究就是以假肢的仿生性作为主要方向，目前，智能仿生假肢是康复工程领域的国际研究热点。本项目基于人体肢体运动的神经生理电活动现象，以运动意识对应的脑电信号与残肢肌肉活动所产生的肌电信号为出发点，研究了脑电信号与肌电信号相结合实现多自由度假手运动控制的方法。针对头皮脑电和表面肌电信号微弱、背景噪声强的特点，研究了基于小波消噪、独立分量分析、盲源分离等算法的信号消噪和复原方法。考虑到生物电信号非平稳性和随机性，在分析对比多种特征提取方法的基础上，采用了非线性动力学参数来提取信号特征。以基本尺度熵表达肌电信号的特征、排列组合熵表示脑电信号特征，形成多维特征向量，运用支持向量机和D-S证据推理实现融合和动作分类决策。本项目还研制了一款具有腕内旋、腕外旋、伸腕、屈腕、展拳和握拳动作的电动假手，搭建了脑电/肌电信号采集与分析平台，所述算法在研制的假手上进行了实验，六种待识别动作模式成功分类达到87%以上，克服了单一肌电信号控制多自由度肌电假手中存在的动作识别率较低的问题。本研究为电动机械假手提供了一种新的仿生控制方法。

中文关键词： 假肢；脑电；肌电；控制

英文摘要： Physical disabilities desire to dominate their own body like the normal persons, as a consequence, the research on the high-performance electric prosthesis is mainly oriented to study the bionic properties of the prosthetic limbs. At present, intelligent bionic prosthesis is an international hotspot in the field of the rehabilitation engineering. Based on the phenomenon of the neural physiological electrical activity of human limb movement, this project is tempted to work out the method to achieve the control of the prosthetic hand's motion under multiple degrees of freedom combining electroencephalogram (EEG) signals and electromyography (EMG) signals, which EEG signals corresponded to the physical movement consciousness and EMG signals generated by the residual limb muscle activities. For the scalp EEG and surface EMG signal is weak, and strong background noise, methods of signal de-noising and reconstruction are studied based on the algorithm such as wavelet de-noising, independent component analysis, blind-source separation and so on. Simultaneously, considering the randomness and non-stationary of biological electrical signals, the nonlinear dynamic parameters are adopted to extract the signals' features on the basis of the analysis and comparison of the multiple features extraction method. In this project, it's reasonable to get the multidimensional eigenvector from the features of EMG signals expressed by the basic scale entropy and that of EEG signals expressed by the permutation and combination entropy, which can be used to achieve the data fusion and the motion classifying decisions by the support vector machine and the Dempster-Shafer evidence theory. What's more, there is a kind of electric prosthetic hand developed, the motions of which includes wrist extorsion, wrist intorsion, wrist flexion, wrist extension, fist clenching and fist unfolding. As well, a platform to acquire and analyze EEG and EMG signals is put up in this project and the algorithm mentioned is experimented to control the electric prosthetic hand. The rate of successful classification of the six movements is over 87%, which overcomes the low recognition rate existing in controlling the multiple degree of freedom EMG prosthetic hand by the single EMG signals. A new bionic control method is provided for the mechanical electric prosthetic hand in this project.

英文关键词： prosthesis; electroencephalogram; electromyography; control