The IBM HeadTracking Pointer: Improvements in vision-based pointer control

Abstract

Vision-based head trackers have been around for some years and are even beginning to be commercialized, but problems remain with respect to usability. Users without the ability to use traditional pointing devices – the intended audience of such systems – have no alternative if the automatic bootstrapping process fails. There is room for improvement in face tracking, and the pointer movement dynamics do not support accurate and efficient pointing. This paper describes the IBM HeadTracking Pointer, a system which attempts to directly address some of these issues. Head gestures are used to provide the end user a greater level of autonomous control over the system. A novel face-tracking algorithm reduces drift under variable lighting conditions, allowing the use of absolute, rather than relative, pointer positioning. Most importantly, the pointer dynamics have been designed to take into account the constraints of head-based pointing, with a non-linear gain which allows stability in fine pointer movement, high speed on long transitions and adjustability to support users with different movement dynamics. User studies have identified some difficulties with training the system and some characteristics of the pointer motion that take time to get used to, but also good user feedback and very promising performance results.

Keywords:

• Head tracking
• pointer control
• computer vision