چكيده به لاتين
Abstract:
EEG signals have essential and important information about brain and neural disease. The purpose of this study is classifying 2group (7healthy volunteers and 5M.S patient). EEG signals are recorded when users are doing two task. These are attention tasks which one of them is based on color luminance changing and the other is based on direction changing. In addition to EEG signals, we want to analysis EEG sub-bands. So EEG sub-bands were extracted. After recording and preprocessing, for EEGs and sub-bands, state space is reconstructed by time delay embedding method and embedding parameters are determined. Then we use nonlinear features include Lyapunov Exponents, Approximate Entropy, Sample Entropy, Hurst Exponent, Higuchi and Katz fractal Dimensions and L-Z complexity for feature extracting. In order to reduce the feature numbers feature selection is done by T-test criterion. For direction and color luminance based tasks, classification performances are 93.08% and 79.79% respectively (both in optimal feature numbers).
Keywords: Electroencephalogram (EEG), Multiple Sclerosis (MS), Attention Task, Nonlinear Features, Support Vector Machine (SVM)