A bag of words model for improving automatic stress classification

Abstract

Most of the existing stress assessment frameworks rely on physiological signals measurements (EEG, ECG, GSR, ST, etc.), which involve direct physical contact with the patient in a medical setup. Present technologies rely on capturing moods and emotions through remote devices (cameras), further processed by computer vision and machine learning techniques. The proposed work describes a method of automatic stress classification where stress information is modeled based on pupil diameter non-intrusive measurements, recorded by an eye tracking remote system. The signal extracted from the pupil Dataset has been processed using the Bag-Of-Words model, with a SVM classification and results have been compared to similar experiments in order to validate the applicability and consistency of the Bag-Of-Words model on stress assessment and classification.