Approximated Match Swiping: Exploring more Ergonomic Gaze-based Text Input for XR

In augmented and virtual reality, hands-free text input is still a significant challenge. Though some virtual keyboard systems exist, they often require the use of controllers, suffer from the Midas touch problem, or require significant activation times to allow for the selection of keys. To help address this problem, we present a system called Approximated Match Swiping (AMSwipe for short), a pure gaze-based text input system that allows for efficient typing in virtual environments. This system builds on prior eye-based path/swipe systems for word recognition and selection in which users look at the keys on a virtual keyboard in succession to trace out a word path. Unlike other systems such as reverse-crossing, our method makes use of the spacebar to start or stop input, and the user can chain words together with additional brief glances at the spacebar, allowing for faster transitions between words. This system works for virtual and augmented reality as well as desktop and mobile systems. In an experiment testing our design alongside other well-known input systems, participants achieved an average raw typing speed of 7.81 words per minute (WPM) with under five minutes of practice, with an expert user reaching an average of 23 WPM. We compared our technique with two others, including Dwell and reverse-crossing, which had similar novice typing speeds of 8.19 WPM and 6.40 WPM, respectively.