Studying stress via skin conductance in Virtual Reality environments.

Studying stress via skin conductance in Virtual Reality environments. Our skin could be a valid support in the study of emotions in stressful situations, even when the face may hide some of them. In literature, the most common way to elicit an emotional response is to provide audiovisual stimuli, assuming the fact that these could give physiological response. The analysis of galvanic skin response (GSR) or electrodermal activity (EDA), or skin conductance (SC), is one of the most sensitive methods to measure arousal. Indeed, GSR is a valid biomarker for stress, and is closely linked to the arousal level. GSR arises from the activation of the sweat glands, i.e. the sweating of fingers, handpalms and feet is a response to emotional stimulation. This thesis project starts from an experiment made at 3D Lab of the Politecnico di Torino that saw thirty-nine participants subject to emotional monitoring during the navigation of two ad hoc designed Virtual Reality environments, with the goal of analysing the skin conductance as an indicator of the stress and arousal levels. A Shimmer3 GSR+ device was used to collect the electrodermal activity during the experiment. The instrument is equipped with two sensors connected to each participant to the index and middle finger of the left hand for EDA collection, and a sensor connected to the ear lobe for heart rate (HR) collec-tion. At the end of each session, participants filled out questionnaires to assess their level of stress and arousal during the experiences. In addition to EDA, the study of the emotional load was also conducted via electroencephalography (EEG), made possible to a helmet (EMOTIV Epoc X), for the collection of the electrical signals of the brain. The experiment is divided in two phases related to the navigation in two Virtual Reality environ-ments, each one consisting of eight tasks. During the first experience, the intent was to induce a stress load given by the shortage of time/knowledge/salvation/alternatives/control (the 'stressors') to the posed problems, while the second phase was less stressful thanks to the alleviation of the stressors. The EDA signals were processed and divided into tonic and phasic components with the use of a Matlab Toolbox (Ledalab). The study focused on five features: the average of the two components, the number of peaks of the skin conductance, the maximum amplitude of the peaks and the average amplitude of the peaks. These features have been statistically analysed with the Shapiro-Wilk test for normality testing and with the Friedman test to evaluate a statistically significant difference be-tween stressful and non-stressful experience. The results achieved by the statistics comfirmed the GSR as a suitable physiological signal for the study of arousal and stress, as the parameters produced by the stressful environment are, on aver-age, greater than those detected in the non-stress phase. The non-parametric Friedman test has led to significant differences between the tonic and phasic components even more relevant. Moreover, the thesis focuses on the training of two classifier, the SVM and KNN, to achieve the divi-sion between "stress" and "non-stress", and between "high arousal" and "low arousal" states. The purpose of the classification is to investigate the match between the questionnaire responses and the EDA indicator.