Automated Behavioral Modeling and Pattern Analysis of Children with Autism in a Joint Attention Training Application: A Preliminary Study

摘要

Although recent research works have highlighted and demonstrated the applicability of the activity and behavioral pattern analysis mechanisms in offering early windows of opportunities in the assessment and intervention for individuals with autism spectrum disorder (ASD), the computational cost and sophistication of such behavioral modeling systems might prevent these automatic and semi-automatic systems from deploying, which might in turn restrict its actual use. As such, in this paper, we proposed an easily deployable automatic system to train joint attention (JA) skills, characterizing and evaluating JA and reciprocity patterns (i.e. the frequency and degree of reciprocity, initiating and responding to JA bids). Our proposed approach is different from most of earlier attempts in that we do not capitalize the sophisticated feature-space construction methodology; instead, the simple designs and in-game automatic data collection offers hassle-free benefits for such individuals as special education teachers and parents to use in both classrooms and at homes.