The feature extraction of ERD/ERS signals based on the wavelet package and ICA

For the feature extraction of motor imaginary EEG (electroencephalography) in the study of brain-computer interface(BCI), a method of extracting EEG features based on wavelet package combined with ICA (Independent component analysis) was adopted to extract the signals produced by imaginary movement, event related desynchronization or event related synchronization (ERD/ERS). First, in order to eliminate the statistical correlation between different EEG rhythms, the EEG signal was decomposed into five levels by wavelet packet. At the same time some sub-band with the characteristics of notable ERD/ERS phenomenon were extracted. Then ICA was separately applied on the feature sub-band components to obtain the μ rhythms and β rhythms corresponding to the ERD/ERS phenomenon. Finally, ERD/ERS coefficient was introduced as a quantity index for the recognition of imaginary movements. The classification results show that the adopted method can significantly enhance the feature information of ERD/ERS produced by imaginary movement. Compared with applying one of two methods separately, the method combined wavelet package and ICA is more efficient to extract feature wave.