基于多类运动想象异步脑—机接口系统的研究

发布时间：2018-07-14 15:58

【摘要】：在头皮采集得到的脑电信号（Electroencephalogram，EEG）是脑细胞电生理活动的整体反映，与人的意识活动状态相关，只要对脑电信号进行分析，就可以识别出不同的意识活动，从而形成一种不依赖于大脑外周神经与肌肉正常输出通道的通讯控制系统，即脑-机接口（Brain-Computer Interface，BCI）。运动想象是指只想象肢体运动而没有进行实际的肢体动作，运动想象产生的脑电信号具有事件相关同步（event-related synchronization,ERS）和去同步（event-related desynchronization, ERD）特征，基于它的脑-机接口系统具有使用者不易疲劳、不依赖外界刺激器、适用人群广、更符合使用习惯的优点而备受关注，是研究热点之一。虽然运动想象脑电信号受到广泛关注，但目前仍然存在很多急需解决的关键问题，如：基于左右手、脚、舌的四类运动想象研究仍停留在离线分析阶段，在线效果还远达不到实际要求。基于左右手的两类运动想象虽然有较好的在线效果，但产生的控制命令十分有限，而且属于同步工作模式，，使用者无法完全自主控制。因此，本文主要针对基于运动想象的在线BCI系统如何进一步提高精度和速度、增加控制自由度和实现异步工作进行研究。本文对四类运动想象脑电信号的采集、处理和异步脑-机接口系统的设计进行了深入研究。采集部分对电极的安放位置、导联方式以及采集实验的具体设计流程进行了阐述，设计并实现了四类运动想象脑电信号的采集；预处理部分采用独立分量分析和FIR数字滤波器分别去除眼电、肌电等干扰，通过比较滤波前后的小波时频图，对滤波效果进行了分析；特征提取部分选用了功率谱估计、小波包分解和希尔伯特黄变换三种算法提取运动想象脑电信号的特征向量，并基于距离准则对特征向量进一步简化，得到最优特征向量；模式识别部分采用一对一法构建多分类支持向量机，并利用遗传算法对其参数进行优化，通过对运动想象脑电信号的特征进行分类实验，比较优缺点，选择出了较为理想的特征提取算法，为实时在线BCI系统分类器的选择提供了依据；最后，结合Alpha波和运动想象两种脑电信号的优势，设计控制策略，在LabVIEW平台上建立了异步脑-机接口系统，实现了光标的控制和网页浏览功能。
[Abstract]:Electroencephalograms (EEG) collected on the scalp are an integral reflection of the electrophysiological activity of the brain cells, which is related to the state of human consciousness activity. As long as the EEG signal is analyzed, different conscious activities can be identified. Thus, a communication control system, Brain-Computer Interface (BCI), is formed, which is independent of the normal output channels of peripheral nerve and muscle. Motion imagination refers to the movement of the limbs without actual body movements. The EEG generated by the motion imagination has the characteristics of event-related synchronization and event-related synchronization. The brain-computer interface system based on it has many advantages, such as the user is not easy to fatigue, does not rely on the external stimulator, is suitable for a large number of people, and is more in line with the usage habits, so it is one of the research hot spots. Although the electroencephalogram (EEG) signal of motion imagination has received extensive attention, there are still many key problems that need to be solved. For example, the study of motion imagination based on left and right hand, foot and tongue is still at the stage of offline analysis. Online effect is still far from the actual requirements. Although the two kinds of motion imagination based on the left and the right hand have better online effect, the control commands produced are very limited, and they belong to the synchronous working mode, so the users can not control themselves completely. Therefore, this paper mainly focuses on how to improve the accuracy and speed of online BCI system based on motion imagination, increase the control degree of freedom and realize asynchronous work. In this paper, the acquisition, processing and design of asynchronous brain-computer interface system for four kinds of motion imagination EEG signals are studied. In the collection part, the location of the electrode, the lead mode and the specific design flow of the collection experiment are described, and the collection of four kinds of motion imaginary EEG signals is designed and realized. In the preprocessing part, independent component analysis (ICA) and Fir digital filter are used to remove EMG and EMG respectively. The filtering effect is analyzed by comparing the wavelet time-frequency images before and after filtering, and the power spectrum estimation is used in the feature extraction part. Wavelet packet decomposition and Hilbert-Huang transform are used to extract the eigenvector of the motion imaginary EEG signal, and the optimal eigenvector is obtained by further simplification of the eigenvector based on the distance criterion. In the part of pattern recognition, a one-to-one method is used to construct multi-classification support vector machine, and its parameters are optimized by genetic algorithm. The advantages and disadvantages are compared by classifying the characteristics of motion imaginary EEG signals. An ideal feature extraction algorithm is selected, which provides a basis for the selection of real-time online classifiers. Finally, combining the advantages of Alpha wave and motion imagination, a control strategy is designed. An asynchronous brain-computer interface system is established on LabVIEW platform, which realizes cursor control and web browsing.
【学位授予单位】：天津职业技术师范大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP334.7;TN911.7

【参考文献】