Generating Efficient Feedback and Enabling Interaction in Virtual Worlds for Users with Visual Impairments

Abstract

Virtual worlds have become very popular in the last decade and now these simulated environments have tens of millions of active users worldwide. Users benefit from virtual worlds in many different areas including social interaction, gaming, education and business. However, for users with visual impairments, accessing virtual worlds and benefiting from its facilities is rather challenging because of a number of important accessibility barriers. Our previous research effort, TextSL, has overcome perhaps the biggest barrier by providing a screen-reader accessible interface for individuals with visual impairments. Nevertheless, problems arising because of the limitations of atext-based desktop interface, challenges in textually describing a virtual scene and lack of reliable meta-data stand in front of our ultimate aim of enabling accessibility in virtual worlds. This thesis presents our contribution to virtual world accessibility by (1) providing a more accessible, portable and feature-rich virtual world interface; (2) proposing a method for generating efficient textual feedback for highly graphical environments; (3) analyzing and evaluating a hybrid virtual object labeling approach that benefits from machine learning and human computation.