Fractal Components From Electroencephalogram ProvideFeatures For Brain Computer Interface

This paper presents an application of nonlinear science to investigate the alterationof electroencephalographic (EEG) signals in two brain states, i.e. idle state and imaginationof movement. Based on Coarse Graining Spectrum analysis (GCSA) we extract fractalcomponents from the recorded EEG signal in the frequency domain. The aim of this work isto extract reliable features based on random fractal components to apply as features in aBrain Computer Interface (BCI). Data recorded from six subjects for a BCI system is used toevaluate the proposed method. The result derived from Soft Margin Support Vector Machineclassification shows that there is significant discrepancy in spectral components that could bea powerful method in investigating brain activities.