Automated and Adaptive Feature Extraction for Brain-Computer Interfaces by using Wavelet Packet

An automated and adaptive feature extraction method is discussed in this paper. The method is based on the wavelet packet transform (WPT) and used to extract features of electroencephalogram (EEG) signals for brain computer interfaces (BCIs). The idea is to employ the best basis algorithm to select the most appropriate wavelet and the best wavelet packet basis automatically. Meanwhile, both the selected wavelet and the selected basis are adaptive to each EEG channel and each subject. The effectiveness of the method is verified by discriminating three different motor imagery tasks of six subjects