Comparing selection mechanisms for gaze input techniques in head-mounted displays. Issue 139 (July 2020)

• Record Type: Journal Article
• Authors: Esteves, Augusto; Shin, Yonghwan; Oakley, Ian
• Is Part Of: International journal of human-computer studies. Issue 139(2020)
• Publication Date: 2020-07
• Subjects: Hands-free input -- Head pointing -- Head-mounted display -- Virtual-reality -- Augmented-reality -- Gaze input -- Motion matching; Human-machine systems -- Periodicals; Systems engineering -- Periodicals; Human engineering -- Periodicals; Human engineering; Human-machine systems; Systems engineering; Periodicals; Electronic journals; 004.019
• DOI: 10.1016/j.ijhcs.2020.102414
• Languages: English
• ISSNs: 1071-5819

Abstract

Highlights: The clicker and dwell were the most popular selection mechanisms across the studies. The clicker led to fewer errors, target re-entries, and was less taxing than on-device input. Dwell was faster, more accurate, and more stable than gesture and speech. Motion matching produced an equivalent amount of errors to the clicker and was faster than dwell. Edge targets take no longer to select than center targets with motion matching (Δ32ms). Abstract: Head movements are a common input modality on VR/AR headsets. However, although they enable users to control a cursor, they lack an integrated method to trigger actions. Many approaches exist to fill this gap: dedicated "clickers", on-device buttons, mid-air gestures, dwell, speech and new input techniques based on matching head motions to those of visually presented targets. These proposals are diverse and there is a current lack of empirical data on the performance of, experience of, and preference for these different techniques. This hampers the ability of designers to select appropriate input techniques to deploy. We conduct two studies that address this problem. A Fitts' Law study compares five traditional selection techniques and concludes that clicker (hands-on) and dwell (hands-free) provide optimal combinations of precision, speed and physical load. A follow-up study compares clicker and dwell to a motion matching implementation. While clicker remains fastest and dwell most accurate, motion matching may provide a valuable compromise between these two poles.