SkiMR: Dwell-free Eye Typing in Mixed Reality

We present SkiMR: a dwell-free eye typing system that enables fast and accurate hands-free text entry on mixed reality headsets. SkiMR uses a statistical decoder to infer users’ intended text based on users’ eye movements on a virtual keyboard. It does not rely on dwell timeouts for key selections, which enables it to be faster than traditional eye typing. We study this dwell-free eye typing system, deployed on a HoloLens 2, in two studies. In the first study (n = 12) we show that dwell-free eye typing results in a significantly faster text entry rate compared to traditional dwell-based eye typing with word prediction support, and a hybrid dwell-free method that uses dwell timeouts to delimit word entry. Based on the insights from the first study we evaluate the feasibility of a refined system in a more realistic composition task and use an interaction mechanism that provides real-time predictions during dwell-free eye typing. The second study (n = 16) demonstrates that this final system allow users to compose original text at 12 words per minute with a corrected character error rate of 1.1%. Overall, this work demonstrates the high potential for fast and accurate hands-free text entry using dwell-free eye typing for mixed reality headsets.