The purpose of this paper is to select important features to detect attention stage using non-linear and frequency band features from EEG(Electroencephalographic). Attention is one of brain cognitive activities and has correlation with learning ability, accidents and diseases. In this paper, 13 features were extracted by non-linear and frequency band analysis. The features are used as inputs for GA(Genetic algorithm) based SVM(Support vector machine) to classify attention stage and select

more »

... al features. The classification accuracy (87%) of GA-SVM is higher than that of SVM alone (83%) and use of both non-linear and frequency band feature sets improves the classification accuracy in comparison non-linear or frequency band feature sets alone. The optimized selection parameters are NRP(Number of recurrence plot),  and . The reduced set of parameters can be used for a real-time attention classification system.