论文部分内容阅读
在脑-机接口(BCI)中,为了提高小样本学习运动想象脑电信号(EEG)的分类识别准确率,本文提出一种基于相关系数分析的特征选择方法。针对2005年BCI竞赛数据集Ⅳa中5位样本数据,通过短时傅里叶变换(STFT)和相关系数的计算,降低了原始EEG信号的维数,然后进行共空间模式(CSP)特征提取与线性判别分类器(LDA)的分类识别。仿真实验表明,运用相关系数分析的分类性能远远优于未经特征优化的结果,与支持向量机(SVM)的特征优化算法相比,相关系数分析方法能更好地选择导联参数,提高分类识别准确率。
In BCI, in order to improve the classification accuracy of EEG, a novel feature selection method based on correlation coefficient analysis is proposed in this paper. According to the five sample data of 2005 BCI dataset Ⅳa, the dimension of the original EEG signal is reduced by short-time Fourier transform (STFT) and the correlation coefficient, and then the feature extraction of CSP Classification and Recognition of Linear Discriminant Classifier (LDA). Simulation results show that the classification performance using correlation coefficient analysis is far superior to the results without feature optimization. Compared with the feature optimization algorithm based on Support Vector Machine (SVM), the correlation coefficient analysis method can better select the lead parameters and improve Classification recognition accuracy.