Hossein Moghimi (Birmingham)

Adaptive Virtual Environments - A Psychophysiological Feedback HCI System Concept
This project aims to design an adaptive dynamic virtual environment, capable of responding to human emotions. Based on the development of a Valence-Arousal-Dominance “Circumplex” (model of emotions), a controllable affective virtual medium (a computer game capable of evoking multiple emotions on the users) has been constructed. The project included five phases: 1) Designing a generic game scenario which can incorporate a large set of variables, with potential variable impact on the users’ emotional experience; 2) Using an online survey with 35 participants to assess the potential emotional impact of each variable; 3) Designing games with combined variables to maximise their emotional effect. The results were validated using additional 68 participants, who played and emotionally rated their experiences. 4) A physiologically-based experiment has been executed, in which the EEG, GSR and Heart Rate of 30 male and female gamers have been recorded during exposure to the most powerful affective environments, identified in the earlier study. A physiological database, with corresponding processed game events and self-reported emotional experiences, has been constructed to be used in the design and evaluation of an affective computing system. 5) More than 700 physiological features have been extracted from the training database, while only minority of them have been selected, to be used in the classification process, using a variable selection algorithm. The selected features have been used to train an affective computing system, using 3 different classification techniques; K-Nearest Neighbour (KNN), Linear Discriminant Analysis (LDA) and Support Vector Machines (SVM). The performance of the different classification techniques (each with various settings) have been compared, using a 10-Fold Cross validation. The best classifier has been able to identify subjects’ emotions, with 92% accuracy, using KNN technique while employing only 4 features.