The potential of fusing computer vision and depth sensing for accurate distance estimation

Authors:
Jakub Dostal;Per Ola Kristensson;Aaron Quigley
Affiliations:
University of St Andrews, St Andrews, Fife, UK;University of St Andrews, St Andrews, Fife, UK;University of St Andrews, St Andrews, UK
Venue:
CHI '13 Extended Abstracts on Human Factors in Computing Systems
Year:
2013

Citing 5
Cited 2

Robust Real-Time Face Detection

International Journal of Computer Vision
Interactive public ambient displays: transitioning from implicit to explicit, public to personal, interaction with multiple users

Proceedings of the 17th annual ACM symposium on User interface software and technology
Lean and zoom: proximity-aware user interface and content magnification

Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
Proxemic interaction: designing for a proximity and orientation-aware environment

ACM International Conference on Interactive Tabletops and Surfaces
Seamless interaction in space

Proceedings of the 23rd Australian Computer-Human Interaction Conference

Multi-view proxemics: distance and position sensitive interaction

Proceedings of the 2nd ACM International Symposium on Pervasive Displays
SpiderEyes: designing attention- and proximity-aware collaborative interfaces for wall-sized displays

Proceedings of the 19th international conference on Intelligent User Interfaces

Quantified Score

Hi-index	0.00

Visualization

Abstract

Accurately inferring the distance between the user and the interface enables the design of a variety of proximity-aware user interfaces. This paper reports our work-in-progress on designing a toolkit called SpiderEyes that will enable systems to accurately estimate the distance between the user and the interface by fusing computer vision with depth sensing. Potential advantages of this approach include increased accuracy and the ability to reliably estimate the user's distance to the interface even when the user is far from the sensor (up to five metres). We evaluated the feasibility of this approach in a controlled experiment and found that it is likely to yield distance estimations with less than a 10 cm estimation error when users are between 50 cm and 5 metres away from the system.