Feature extraction and classification of four-class motor imagery EEG data

Due to the low information transfer rate and low recognition accuracy in brain computer interface (BCI), feature extraction and classification of multi-channel four-class motor imagery for electroencephalogram(EEG)-based BCI was investigated. Optimum filtering band was obtained for power spectral analysis of four-class motor imagery and resting EEG. Then, the PW-CSP, Hilbert transformation and normalization were used to extract the feature of EEG data. Classification was divided into two steps, the first step was arithmetic summation and threshold comparison, Secondly a single support vector machine (SVM) was applied if the first step failed. The algorithm was simpler than combined SVM, which provided the foundation for on-line application. The experimental results show that the algorithm produces high classification accuracy and less time consumption, moreover, classification result can be further improved at the expense of algorithmic complexity by adjust the threshold.