Computer Vision: From Motion Capture to Tabletop Interaction

Sammanfattning: This thesis presents the work done by the candidate in the domain of Human-Computer Interaction. In particular, the candidate has investigated how techniques in Computer Vision can be used to enable new forms of interaction. The work investigates areas such as motion capture and multi-user, multi-touch interaction for tabletop systems. In the field of motion capture the candidate has investigated the so-called missing markers problem in optical motion capture systems. A first approach to tackling the missing marker problem is presented and evaluated. This approach is based on extrapolation of trajectory information and the retrofitting of such trajectory onto a perfect line or circle. In a second approach, tackling the weak points of the previous one, a more stable solution using a so-called constraint matrix is proposed. The computer vision techniques used in optical motion capture can be extrapolated, adapted and reused to solve tracking problems in, for instance, optically enabled tracking systems for multi-touch interaction. In the domain of tabletop interaction a novel approach to multi-user identification and multi-point interaction is presented along with two applications. The approach makes use of light emitting devices and of the light spot these cast when shined on a surface. The color, size and shape of the light spot enables for a six degrees of freedom device and for user identification. Finally, an emergency management application taking advantage of some of these features is presented.