Abstract. Most text entry methods require users to have physical devices within reach. In many contexts of use, such as around large displays where users need to move freely, device-dependent methods are ill suited. We explore how selection-based text entry methods may be adapted for use in mid-air. Initially, we analyze the design space for text entry in mid-air, focusing on singlecharacter input with one hand. We propose three text entry methods: H4 MidAir (an adaptation of a game controller-based method by MacKenzie et al.[21]), MultiTap (a mid-air variant of a mobile phone text entry method), and Projected QWERTY (a mid-air variant of the QWERTY keyboard). After six sessions, participants reached an average of 13.2 words per minute (WPM) with the most successful method, Projected QWERTY. Users rated this method highest on satisfaction and it resulted in the least physical movement.

Word-gesture keyboards enable fast text entry by letting users draw the shape of a word on the input surface. Such keyboards have been used extensively for touch devices, but not in mid-air, even though their fluent gestural input seems well suited for this modality. We present Vulture, a word-gesture keyboard for mid-air operation. Vulture adapts touch based word-gesture algorithms to work in midair, projects users' movement onto the display, and uses pinch as a word delimiter. A first 10-session study suggests text-entry rates of 20.6 Words Per Minute (WPM) and finds hand-movement speed to be the primary predictor of WPM. A second study shows that with training on a few phrases, participants do 28.1 WPM, 59% of the text-entry rate of direct touch input. Participants' recall of trained gestures in mid-air was low, suggesting that visual feedback is important but also limits performance. Based on data from the studies, we discuss improvements to Vulture and some alternative designs for mid-air text entry.

Figure 1. (a) WatchSense enables on-and above-skin input on the back of the hand (BOH) through a wrist-worn depth sensor. (b) Our prototype mimics a smartwatch setup by attaching a small depth camera to the forearm. (c) It tracks the 3D position of fingertips, their identities, and touch on the BOH in real-time on consumer mobile devices. This enables a combination of mid-air and multitouch input for interactive applications on the move.

The size of information spaces often exceeds the limits of even the largest displays. This makes navigating such spaces through on-screen interactions demanding. However, if users imagine the information space extending in a plane beyond the display's boundaries, they might be able to use the space beyond the display for input. This paper investigates Off-Limits, an interaction concept extending the input space of a large display into the space beyond the screen through the use of mid-air pointing. We develop and evaluate the concept through three empirical studies in onedimensional space: First, we explore benefits and limitations of off-screen pointing compared to touch interaction and mid-air on-screen pointing; next, we assess users' accuracy in off-screen pointing to model the distance-to-screen vs. accuracy trade-off; and finally, we show how Off-Limits is further improved by applying that model to the naïve approach. Overall, we found that the final Off-Limits concept provides significant performance benefits over onscreen and touch pointing conditions.

No abstract