Motor/Imaginary movement classification based on Fractal Dimension FD and event-related desynchronization ERD

Until recently, the dream of being able to control one's environment through thoughts had been in the realm of science fiction. However, the advance of technology has brought a new reality: Today, humans can use the electrical signals from brain activity to interact with, influence, or change their environments. The emerging field of brain-computer interface (BCI) technology may allow individuals unable to speak and/or use their limbs to once again communicate or operate assistive devices for walking and manipulating objects. In this study, a PhysioNet data set have been used to create a BCI system, by investigating 1527 EEG signals during real and imagination movement tasks in time domain using Fractal Dimension (FD), and in frequency domain using Event related desynchronisation (ERD). Different classifiers such as SVM, KNN and TREE have been applied with various combination of EEG channels. Moreover, the problem of selecting channels group have been discussed by proposing a new technique to select the optimal channels combination for each task and each classifier. It has been found that the optimal channels selection technique is significantly able to improve the performance of each classifier. With accuracy equal to 98% for classifying between real and imagination of hand movement (se= 97%, sp=99%, AUC=0.99) obtained by using ERD with SVM. While the accuracy of classifying between the imagination of hand and feet movements equal to 91% (se=87% , sp=86% , AUC=0.93 ). This study proposed different models of BCI system in order to select the best one with the highest accuracy to help removing the human communication boundaries and to improve the quality of the life for peop