Lisa
Elkin,
Master’s
candidate
David
R.
Cheriton
School
of
Computer
Science
We study a cuboid tangible pen-like input device similar to Vogel and Casiez’s Conte. A conductive 3D-printed Conte device enables touch sensing on a capacitive display and orientation data from an enclosed IMU reliably distinguishes all 26 corners, edges, and sides. The device’s size is constrained by hardware required for sensing. We evaluate the impact of size form-factor on manipulation times for contact-to-contact transitions. A controlled experiment logs manipulation times performed with three sizes of 3D printed mock-ups of the device. Computer vision techniques reliably distinguish between all 26 possible contacts, and a resistive touch sensor provides accurate timing information. In addition, a transition to touch input is tested, and a mock-up of a digital pen is included as a baseline comparison.
Results show larger devices are faster, contact-to-contact transition time increases with distance between contacts, but transitions to barrel edges can be slower than some end-over-end transitions. A comparison with a pen-shaped baseline indicates no loss in transition speed for most equivalent transitions.
Based on our results, we discuss ideal device sizes and improvements to the simple extruded-rectangle form-factor. Subsequently, we evaluate learning and recall of commands located on physical landmarks on the exterior of a 3D tangible input device in comparison with a 2D spatial interface. Each of the 26 contacts is a physical spatial landmark on the exterior of Conte. A pilot study compares command learning and recall for Conte with a 2D grid interface, using small and large commands sets. To facilitate novice learning, an on-screen model of Conte replicates the physical device’s orientation and displays icons representing commands on the corresponding landmarks.
Results show there is likely no difference between 2D and 3D spatial interface recall for a small command set and high recall is possible with large command sets. Applications illustrating possible use cases are discussed as well as possible improvements to the on-screen guide based on our results.