A fractal approach is employed for the brain motor imagery recognition and applied to brain computer interface (BCI).
The fractal dimension is used as feature extraction and SVM (Support Vector Machine) as feature classifier for on-line
BCI applications. The modified Inverse Random Midpoint Displacement (mIRMD) is adopted to calculate the fractal
dimensions of EEG signals. The fractal dimensions can effectively reflect the complexity of EEG signals, and are related
to the motor imagery tasks. Further, the SVM is employed as the classifier to combine with fractal dimension for
motor-imagery recognition and use mutual information to show the difference between two classes. The results are
compared with those in the BCI 2003 competition and it shows that our method has better classification accuracy and
mutual information (MI).