基于运动想象的脑—机接口频带优化及分类算法研究

发布时间：2018-03-13 08:41

本文选题：脑-机接口　切入点：共空域模式　出处：《南昌大学》2012年硕士论文　论文类型：学位论文

【摘要】：脑-机接口是一种能将大脑意识信号转化成外部输出命令的技术。在基于运动想象的脑-机接口系统中,共空域模式(Common Spatial Pattern, CSP)是一种成功的算法。它的优势在于能设计一种最优的空域滤波器,通过空域滤波可以提取脑电信号(electroencephalogram, EEG)的空域特征。但是共空域模式算法的性能很大程度上也取决脑电信号的频域信息。因此如何提取脑电信号的最优频带就显得尤为重要。为了有效地解决脑电信号频带优化选择的问题,本文提出了两种基于CSP的频带优化选择算法。第一种算法是小波包系数加权的方法,适用于二进制(即二分类)脑-机接口。该算法在小波包分解对频带划分的理论基础上,通过对小波包系数加权来实现最优频带的选择；第二种方法适用于多分类脑-机接口,是滤波器组与特征选择相结合的频带优选方法。该算法使用滤波器组将原始的脑电信号分解为为多个子带信号,使用CSP算法提取每个子带信号的特征,通过特征选择实现频带优化。在离线分析实验中,两种算法都取得了不错的分类效果。相对于宽带方法,这两种算法都使分类识别率得到较大幅度的提升。
[Abstract]:Brain-Computer Interface (BCI) is a technology that converts brain consciousness signals into external output commands. In brain-computer interface systems based on motion imagination, Common Spatial pattern (CSP) is a successful algorithm, which has the advantage of designing an optimal spatial filter. The spatial feature of EEG can be extracted by spatial filtering. However, the performance of the common spatial pattern algorithm also depends on the frequency domain information of EEG to a great extent. Therefore, how to extract the optimal frequency band of EEG signal is obvious. In order to effectively solve the problem of optimal selection of EEG frequency band, In this paper, two optimal selection algorithms based on CSP are proposed. The first method is the weighted method of wavelet packet coefficient, which is suitable for binary (i.e. two-classification) brain-computer interface. This algorithm is based on the theory of wavelet packet decomposition of the frequency band partition. The selection of optimal frequency band is realized by weighted wavelet packet coefficient. The second method is suitable for multi-classification brain-computer interface. The algorithm uses the filter bank to decompose the original EEG signal into multiple sub-band signals, and uses the CSP algorithm to extract the features of each sub-band signal. In off-line analysis experiments, the two algorithms have achieved good classification results. Compared with the wideband method, both of these two algorithms can improve the classification recognition rate by a large margin.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP334.7

【参考文献】