关键词: 脑电信号, 脑-机接口, 小波方差, 线性分类, 分类时间

Abstract: A wavelet variance-based method to extract feature was adopted for a BCI task of two different imagined movements, that is, the imaginary left and right hand movements. Firstly, the computational methods of wavelet transform and wavelet variance were discussed in depth, then the features of EEG signals from the electrodes C3 and C4 using the variances of wavelet coefficient were extracted based on the ERD/ERS phenomenon. Finally, they were classified by using a most simple linear classifier, and classification accuracy and time interval were taken as evaluation criteria for BCI system. The results showed that the maximum classification accuracy was 85% and the best time interval for classification was 4~6.5 s. On precondition of guaranteeing the accuracy, the method for feature extraction and classification described is more efficient and simpler than BCI competition and others, which can be regarded as a good reference.