Classification of Mental Arithmetic States Using Single Electrode

Mental stress has invariably become a part of every one's life today. According to the American institute of stress approximately 35% of people are under stress considering 143 countries. Americans average out to 55% of the population under stress. Early detection shall help person from further slipping into depression. The dataset used in this study is physionet's EEGMAT dataset. 36 subjects participated in the study. In this paper we propose method to classify the mental work load conditions and rest state using only time domain features. We use wavelet decomposition to extract the alpha band of EEG which shows differentiation between the two mental states specially in the occipital area. We select only 1 electrode “O1” form the occipital area for the analysis. Using various machine learning models like SVM, KNN, decision trees and neural networks we try to classify the mental arithmetic state and the rest state of the brain. The maximum accuracy obtained is 100%.