ANALYSING EUSTRESS AND DISTRESS USING A NEUROPHYSIOLOGICAL COMPUTATIONAL MODEL OF AFFECT (NCMoA)

Abstract

Most contemporary research on stress focuses on its negative appraisal (distress) while its benefits (eustress) have relatively been ignored or relied on self-reported data. These constructs, known as ‘core affect’, can be represented on a two-dimensional (x-y) axis where the x-axis (valence) depicts positive to negative emotions and the y-axis (arousal) depicts drowsiness to excitedness. The first phase of this study examines the relationship between stress appraisal and emotional states in an academic environment. Feedback to an online Google Forms survey consisting of various psychological instruments for assessing stress and affect, was collected. Based on significant correlations between scores for the stress questionnaires and scores for the affective scale, prediction equations were derived. Mean squared error (MSE) performance evaluation showed that the model for Academic eustress had the lowest loss function while the model for ADES eustress had the lowest percentage error. In the second phase of the study, electroencephalography (EEG) signals were collected from the prefrontal cortex region of participants while they viewed happy, fear, calm and sad emotive images from the International Affective Picture System (IAPS) database. Bandpass filtering was used to separate signals according brainwaves while Independent Component Analysis (ICA) was used for artefact removal. Mel Frequency Cepstral Coefficient (MFCC) features are then extracted from brainwaves and fed into a Multilayer Perceptron (MLP) that classifies emotion based on valence and arousal. A memory test was then used to validate the model, with a set accuracy threshold. Participants below the threshold were excluded, with the final model called the neurophysiological computational model of affect (NCMoA), which is used to extract valence and arousal from EEG. Pearson correlation analysis is used to assess the relationship between extracted valence and arousal scores and stress questionnaire scores. Prediction equations were derived, with the model for ADES eustress having the lowest loss function and the model for PSS-10-C having the lowest percentage error. These valence and arousal are extracted from EEG signals of participants as they solved arithmetic tasks of low, medium and high relative difficulty. The emotion and stress scores showed significant correlations and prediction equations were again derived. ADES distress had the lowest loss function. Finally, ADES questionnaire produced models for core affect during arithmetic tasks. So, it was used to determine the boundaries for the minimum and maximum eustress. In summary, this study sheds light on strategies of coping with stress and recognizing its triggers. This provides a pathway to effective stress management.