Stroke-Gesture Input for People with Motor Impairments: Empirical Results & Research Roadmap

We examine the articulation characteristics of stroke-gestures produced by people with upper body motor impairments on touchscreens as well as the accuracy rates of popular classification techniques, such as the $-family, to recognize those gestures. Our results on a dataset of 9,681 gestures collected from 70 participants reveal that stroke-gestures produced by people with motor impairments are recognized less accurately than the same gesture types produced by people without impairments, yet still accurately enough (93.0%) for practical purposes; are similar in terms of geometrical criteria to the gestures produced by people without impairments; but take considerably more time to produce (3.4s vs. 1.7s) and exhibit lower consistency (-49.7%). We outline a research roadmap for accessible gesture input on touchscreens for users with upper body motor impairments, and we make our large gesture dataset publicly available in the community.