ICA+CSSD的脑-机接口分类

针对复杂的脑电问题，介绍了一种对单次思维脑电信号提取、分类的方法。该方法的主要思想是将独立成分分量和共空域子空间分解方法以及支持向量机学习方法结合起来，用于提取脑电信号特征。该方法分别被用于BCI Competition 2003 Data set IV和BCI Competition III Data set I，正确率分别达到了89%和92%。实验证明独立成分分量算法和共空域子空间分解方法能够很好地结合起来进行脑思维的分类，分类正确率很高，是一种快速、稳定可行的分解方法。
关 键 词 脑－机接口; 脑电信号; 独立分量分析; 支持向量机

Abstract Identification and classification technology plays an important part in study of the brain-computer interface (BCI) system. In this paper, an algorithm is presented to deals with the complex brain signals and extract features and classify single-trial electroencephalogram (EEG). The algorithm combines independent component analysis algorithm and common spatial subspace decomposition with support vector machine to extract features from multi-channel EEG and electro cortico gram (ECoG). This algorithm was applied to the results of data analysis show that the proposed method can classify with high accuracy.
Key words brain-computer interface; electroencephalograms; independent component analysis (ICA); support vector machine