运动想象脑机接口的特征提取与模式分类研究

发布时间：2018-07-08 15:14

本文选题：脑机接口 + 运动想象　；参考：《西安电子科技大学》2012年硕士论文

【摘要】：在基于脑电图(EEG)的脑机接口中，运动想象被认为是最有效的方式之一。在运动想象的准备和执行过程中，，大脑的感觉运动区相关皮层在特定频带和特定时段发生事件相关去同步的现象。利用事件相关去同步的现象可以判别左右手的运动想象。针对基于运动想象的脑机接口，本文提出一种新的时间-频率-空间的特征提取方法，即加强的滤波带宽共同空间模式。在该特征提取方法中，发生事件相关去同步的脑区、频带和时段都被认为是特征提取的重要因素。该特征提取方法将信号分成多个频带和多个时段，对每频带和时段的信号进行共同空间模式(CSP)滤波，再利用两种特征选择方法对所有特征进行自动选择。实验结果表明：基于准确率的特征选择方法比基于互信息的特征选择方法有更高的分类准确率。最后利用脑机接口竞赛数据对于本文提出特征提取方法的有效性进行验证。针对加强滤波带宽共同空间模式选取的特征，本文分别利用线性判别分类器，神经网络分类器以及朴素贝叶斯分类器对其分类效果进行比较。实验结果显示：朴素贝叶斯分类器对于该方法提取的特征具有较好的分类性能。本文属于应用数学与生命科学的跨学科交叉研究。
[Abstract]:In EEG-based brain-computer interface, motion imagination is considered to be one of the most effective methods. In the process of preparation and execution of motion imagination, the sensorimotor cortical associated with the brain occurs event-related desynchronization in a specific frequency band and a specific period of time. The phenomenon of event correlation desynchronization can be used to distinguish the motion imagination of the left and right hand. For the brain-computer interface based on motion imagination, a new time-frequency-space feature extraction method is proposed in this paper, which is the enhanced filtering bandwidth common spatial pattern. In this feature extraction method, the event-related de-synchronous brain regions, frequency bands and time periods are considered to be important factors in feature extraction. The feature extraction method divides the signal into multiple frequency bands and multiple periods. The common spatial mode (CSP) filter is used to filter the signals in each frequency band and time period, and then the two feature selection methods are used to automatically select all the features. The experimental results show that the feature selection method based on accuracy has higher classification accuracy than the feature selection method based on mutual information. Finally, the effectiveness of the proposed feature extraction method is verified by using the BCI contest data. Aiming at the feature of common spatial pattern selection of enhanced filtering bandwidth, the classification effects of linear discriminant classifier, neural network classifier and naive Bayes classifier are compared in this paper. Experimental results show that the naive Bayesian classifier has good classification performance for the features extracted by this method. This paper is a cross-disciplinary study of applied mathematics and life science.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP334.7

【相似文献】