An Ensemble Approach to Classify Mental Stress using EEG Based Time-Frequency and Non-Linear Features

This study proposes a novel method for classifying mental stress before and during arithmetic task using combined nonlinear features and time-frequency domain features extracted from electroencephalography (EEG) signals. 36 participants' EEG signals from 21 channels were used as found in the Physionet database. EEG data were initially divided into several 2s time frames in order to extract five time-frequency domain characteristics and four nonlinear features per channel to carry out the classification task. One artificial neural network (ANN) and six statistical machine learning (ML) models were trained using 80% of the features. The remaining 20% of features were applied to each model for prediction. All the models' predictions were aggregated via an ensemble majority voting method which got an accuracy level of 99.74%.