Research on feature extraction algorithms in BCI

In this paper, wavelet packet algorithm, wavelet entropy algorithm and AR model algorithm were investigated for feature extraction. EEG data of six subjects were analyzed while they performed five different mental tasks. Based on the recognition rate under different mental EEG combination and different subject, it proved that wavelet entropy algorithm had better classification accuracy compared with the other two algorithms. The highest recognition rate is up to 98.48%. The research is valuable and significant in the realization of control and communication based on the mental tasks in BCI.