Brain-Computer Interface System of Motor Imagery Base on HHT and Improved CSP

Human motor imagery tasks evoke kinds of electroencephalogram(EEG) signal.A brain computer interface is a system that can translate the signal of brain for using in communication and control.By comparing mainstream methods of time-frequency analysis,this paper has adopted Hilbert-Huang Transform(HHT) to obtain the marginal spectrum and time domain amplitude spectrum to make the time-frequency analysis of EEG.To further enhance noise ratio of the EEG signal,an improved CSP algorithm has been applied.Combined with time-domain energy analysis,reduced dimension feature by sliding window is extracted to make a further improvement in feature selection.The results of experiment indicate that the method can extract feature and classify effectively.