Authors: Islam A. Fouad; Fatma El-Zahraa M. Labib
Addresses: Biomedical Engineering Department, College of Engineering, MUST University, Giza, Egypt ' Biomedical Technology Department, INAYA Medical College, Riyadh, Saudi Arabia
Abstract: In a Brain-Computer Interface (BCI), the user only concentrates on different mental tasks which activate different functional areas in the brain. This work demonstrates the development of a first BCI system in INAYA Medical College based on electroencephalography pattern recognition. A BCI system recorded the activity of the brain and classified it into two different classes: up and down movements. Both offline and online BCI systems were handled. The offline approach was tested using data set provided by University of Tubingen, while the online approach was designed using real-time EEG data that recorded using five scalp electrodes including the reference electrode. 'Blind Source Separation' by independent component analysis was used to remove the artefacts contaminated signal to get artefact-free signal from which certain features were extracted by applying various techniques in both time and frequency domains. The features were fed to five different types of classifiers where classification accuracy reached 90%.
Keywords: BCI; brain-computer interface; EEG signals; electroencephalograms; pattern recognition; feature extraction; classification; KNN; k-nearest neighbour; classification; minimum distance classifier; artefact removal; blind source separation; BSS.
International Journal of Computer Applications in Technology, 2017 Vol.55 No.2, pp.92 - 111
Received: 25 Nov 2015
Accepted: 18 Feb 2016
Published online: 14 Mar 2017 *