Identification of Human Stress Based on EEG Signals Using Machine Learning

Abstract

Mental health greatly affects human physical health. Mental health can be a source of thinking as well as the response center of all activities. The pressures faced, the burden of thoughts, and food patterns can be a source of human psychological conditions. If the human psychological condition is under stress, it can cause disease. The development of intelligent system technology can take advantage of electroencephalogram (EEG) signals to recognize human mental conditions (stressed and normal). The purpose of this research was to determine the most appropriate method in identifying human psychology (stress and normal) from EEG. Based on the EEG signal taken through the recording of the response signal of the human brain, feature extraction is performed. The features taken are the mean, standard deviation, and MAV (Mean Absolute Value) of each subband, and channel. The total data of respondents studied were 20 people, with 10 normal criteria, and 10 stress. Each of the mean, standard deviation, and MAV features was modeled using the Naive Bayes, SVM, KNN, Backpropagation, Regression Logistics, Deep Learning, ID3 methods. The best method for detecting stress and normal is KNN with 97% accuracy.