Psychophysiological Analysis of Affective States in Human-Computer Interaction for Children with Autism Spectrum Disorders

Abstract

This dissertation addresses the problem of how to make technology-based intervention tools for children with autism spectrum disorders (ASD) affect-sensitive. Two computer-based cognitive tasks are designed to elicit the affective states of liking, anxiety, and engagement. A large set of physiological indices are investigated. Subjective reports on affective states from a therapist, a parent, and the child are collected and analyzed. Therapist-like affective models are designed using Support Vector Machines, which yields 82.9% prediction. The models are applied during affect-sensitive human-robot interaction where the robot adapts its behaviors to detected affective states of a child with ASD. This success motivates the design and evaluation of realistic social interaction scenarios for ASD children using virtual environments. The results support the viability of physiology-based affective computing for future ASD intervention.