Analytical Modelling of Affective Data in Human-Machine Interaction System Designed for Children with Autism Spectrum Disorders

Abstract

Autism Spectrum Disorders (ASD) are a group of developmental disabilities characterized by difficulties in social communication and repetitive or restricted behaviors and interests. According to the latest CDC report, 1 in 54 children in the U.S. has ASD. The estimated annual medical and productivity cost for people with ASD in the US was $268 billion in 2015 and is expected to reach $461 billion in 2025. Given the huge social and economic cost of ASD, early assessment and intervention with the help of technology have attracted public attention as they have shown promising results to help alleviate heterogeneous symptoms in young children with ASD. In this dissertation, we focus on intelligent intervention systems to complement the existing intervention resources.

The specific aim of this dissertation was to design technology-empowered services for the ASD community. This dissertation has contributed towards understanding and modeling of affective data of children with ASD in the human-machine interaction (HMI) process. In particular, this dissertation has contributed in several aspects: 1) Descriptive models were proposed to measure the social performance within HMI. 2) A diagnostic model was created to determine factors for success and failure in HMI, and hereby a prescriptive model to improve the intervention protocol and study timeline for HMI. 3) A predictive model was presented to bridge the gap between social performance within HMI and social performance in real life. 4) A predictive model of anxiety based on physiological data was proposed and validated.

The current work demonstrated the potential for enhancing the welfare of children with ASD through technology-empowered social skill training. It also paved the way for a comprehensive, adaptive, and effective technological solution to assist the development of children with ASD.