基于随机森林算法的高维脑电特征优选

在基于脑电的脑-机接口研究中,脑电信号的分类是较为重要的部分.从多导脑电中得到大量可用于分类的特征,并对这些特征进行优选是研究热点.本文应用多分类器组合的分类树方法和自助法重采样技术,结合随机特征选择,使用随机森林组合分类器方法对想象运动实验中的高维脑电特征进行分析.根据不同特征在生成森林过程中被选用于分枝次数的不同,提出了一种有效的特征优选方法,并在特征优选的基础上可以进行导联的筛选.

High-dimensional EEG features selection based on random forests algorithm

EEG features classification is one of the most important tasks in the research on brain-computer interface. There are more and more feature extraction methods to extract lots of information from the mult-channel EEG signal. In this paper, Random forests algorithm is used to analyze the high-dimensional EEG features from the motor imagery experiment. An effective features selection method based on the times that features were used in the growing process of random trees is proposed. The method enhances the accuracy and speed of classification. Also it is used to select EEG leads in EEG classification.