Studying hemianopia via skin conductance in Virtual Reality environments

This thesis project began with an experiment conducted at the 3DLab of Politecnico di Torino. Sixty volunteer participants engaged in various tasks within three specially designed Virtual Reality environments. The primary objective was to analyze skin conductance (SC), photoplethysmogram (PPG), and electroencephalography (EEG) data. Hemianopia is a loss of vision or blindness in half the visual field, usually on one side of the vertical midline. The most common cause of this damage are stroke, brain tumor and trauma. The experiment involved three distinct Virtual Reality environments: a small market, a house, and a driving simulation. In each environment, various tasks to complete in a predetermined order were assigned to patients. To simulate field of view loss, a post process-layer was applied in Unity, illustrating the contrast between “healthy” individuals and those with the simulated visual impairment. The presence or absence of this visual defect was randomly assigned to the participants. At the end of the experiment a questionnaire was given to every participants. It has been demonstrated that higher immersion through Virtual Reality leads to an increased sense of presence experienced by users in certain environments. It can increase user effects and allow real emotions to be activated [1]. The Shimmer3 GSR+ (Galvanic Skin Response) unit offers connection and preamplification for acquiring one channel of Galvanic Skin Response data (Electrodermal Resistance Measurement – EDR/Electrodermal Activity – EDA). This Unit is designed to measure the electrical conductance of the skin and can also capture an Optical Pulse/PPG (photoplethysmogram) signal. By using the Shimmer ear clip or optical pulse probe it, it can convert this signal to estimate heart rate (HR). The EEG part was performed with a 32-channel wireless EEG head cap system designed for high-density brain monitoring called Emotiv Flex 2. The EDA signals were processed with the use of a Matlab Toolbox called Ledalab and the classifiers with the use a Machine Learning tool in Matlab. The thesis focuses on training two classifiers, SVM and KNN, to differentiate between “healthy” and “simulated hemianoptic” subjects, as well as between “high arousal” and “low arousal” states. The aim of this classification is to explore the correalation between questionnaire responses and the EDA indicator.