Improving Touch Accuracy for People with Motor Impairments

Abstract

Touch-enabled devices such as smartphones, tablets, and interactive kiosks are some of the most pervasive technologies in the world today. Despite the overwhelming popularity of touch input, it presents significant accessibility challenges for millions of people with motor impairing conditions such as cerebral palsy and Parkinson’s disease, and for people who experience motor difficulties as the result of injury or old age. This dissertation presents an ability-based design approach towards creating touch models that can improve the accuracy of touch input for people with motor impairments. I conducted two exploratory studies to understand the touch behaviors of people with motor impairments when interacting on an interactive tabletop computer and a smartphone. These exploratory studies led to the design of two new ability-based touch models: Smart Touch, a user-specific template matching algorithm designed initially for tabletops; and Cluster Touch, a combined user-specific and user-independent touch model that improves touch accuracy on smartphones. I also conducted an exploration of using convolutional neural networks to model a user’s touch behavior to predict intended target locations. This work presents a thorough approach to designing touch models that can support the touch abilities of people with motor impairments. I will demonstrate the following thesis: Ability-based touch models can improve touch accuracy on touch screens compared to native touch sensors or existing statistical models for people with motor impairments and for people in motor-impairing situations.