EEG Signals Processing Based on Fractal Dimension Features and Classified by Neural Network and Support Vector Machine in Motor Imagery for a BCI

In this paper, we evaluated several methods to classify electroencephalogram (EEG) signal recorded from left and right hand during motor imagery. Three subjects (two males and one woman) were volunteered to participate in this experiment. The human brain is a complex and nonlinear system; for this reason we have to try like one. We used some complexity measure techniques based on Fractal Dimension in time and frequency to extract the features patterns in EEG signal Motor Imagery. The algorithms that were selected to get the Fractal Dimension (features extraction) on time Detrended Fluctuation Analysis (DFA), Higuchis Method and on frequency was used Power Spectral Density method. Based on these algorithms we can distinguish between three states relaxing, imagination of right and left hand. After this, these features are classified with two different methods Neural Network (NN) and Support Vector Machine (SVM). Finally, the experimental results are considered to apply in a BCI application to move two robotics hands (left and right hand).