Interaction in Augmented Reality primarily relies on raycast pointing and mid-air touch. An alternative consists of using the non-dominant hand as a touch-sensitive surface, enabling more comfortable, less fatiguing input. AR UI design guidelines have so far discouraged this alternative because of poor hand tracking performance when the hands overlap, favoring touchpads in the air near the hand, rather than on the hand. But significant improvements to the hand tracking capabilities of recent commodity headsets suggest that on-hand pads may now be feasible. We develop an on-hand touchpad prototype and conduct two studies that involve both discrete input and continuous control tasks. The first study compares such on-hand pads to baseline in-air and on-object pads, showing comparable performance despite some limitations in tracking accuracy. The second study quantifies the advantage of on-hand and in-air pads over on-object pads during transitions between touchpad input and other physical hand activities.

TriPad enables opportunistic touch interaction in Augmented Reality using hand tracking only. Users declare the surface they want to appropriate with a simple hand tap gesture. They can then use this surface at will for direct and indirect touch input. TriPad only involves analyzing hand movements and postures, without the need for additional instrumentation, scene understanding or machine learning. TriPad thus works on a variety of flat surfaces, including glass. It also ensures low computational overhead on devices that typically have a limited power budget. We describe the approach, and report on two user studies. The first study demonstrates the robustness of TriPad’s hand movement interpreter on different surface materials. The second study compares TriPad against direct mid-air AR input techniques on both discrete and continuous tasks and with different surface orientations. TriPad achieves a better speed-accuracy trade-off overall, improves comfort and minimizes fatigue.

Inefficient menu interfaces lead to system and application commands being tedious to execute in Immersive Environments. HoloBar is a novel approach to ease the interaction with multi-level menus in immersive environments: with HoloBar, the hierarchical menu splits between the field of view (FoV) of the Head Mounted Display and the smartphone (SP). Command execution is based on around-the-FoV interaction with the SP, and touch input on the SP display. The HoloBar offers a unique combination of features, namely rapid mid-air activation, implicit selection of top-level items and preview of second-level items on the SP, ensuring rapid access to commands. In a first study we validate its activation method, which consists in bringing the SP within an activation distance from the FoV. In a second study, we compare the HoloBar to two alternatives, including the standard HoloLens menu. Results show that the HoloBar shortens each step of a multi-level menu interaction (menu activation, top-level item selection, second-level item selection and validation), with a high success rate. A follow-up study confirms that these results remain valid when compared with the two validation mechanisms of HoloLens (Air-Tap and clicker).

The rapid evolution of multi-display environments (MDEs) has created a vacuum in need of novel input devices to optimize interaction in MDEs. In this paper, we propose TDome, a novel touch-enabled 6DOF input and output device to facilitate interactions in MDEs. TDome offers a private display as output, and multiple degrees of freedom as input by combining touch gestures on the display with physical rotation, roll and translation manipulations of the device. TDome allows versatile interactions that address major MDE tasks, which we illustrate through various proof-of-concept implementations: detect surrounding displays, select one display, transfer data across displays, reach distant displays and perform private interactions. We explore TDome's usability and suitability for MDEs through three user studies. First we explore combined physical+touch gestures from which we discard uncomfortable combinations. We experimentally validate their feasibility and come up with a set of 71 combined gestures that are comfortable and ensure a high success rate, i.e. that can be easily performed and efficiently detected. Finally, we collect user feedback to identify natural mappings between gestures and MDE interactions.

While several techniques offer more than one detailed view in Overview+Detail (O+D) interfaces, the optimal number of detailed views has not been investigated. But the answer is not trivial: using a single detailed view offers a larger display size but only allows a sequential exploration of the overview; using several detailed views reduces the size of each view but allows a parallel exploration of the overview. In this paper we investigate the benefits of splitting the detailed view in O+D interfaces for working with very large graphs. We implemented an O+D interface where the overview is displayed on a large screen while 1, 2 or 4 split views are displayed on a tactile tablet. We experimentally evaluated the effect of the number of split views according to the number of nodes to connect. Using 4 split views is better than 1 and 2 for working on more than 2 nodes.