B. Fröba and C. Küblbeck (Germany)
Gesture Recognition, Hand Segmentation, Real-Time Vi sion
In this paper we describe our work in progress on a novel approach to real time hand gesture detection in arbitrary grey level video streams and it’s application to a human computer interface. The hand gestures used here con sist of simple pointing gestures to move a mouse pointer and gestures to emulate click operations and releasing the mouse pointer. A detection algorithm is designed for rapid hand-background segmentation which applies a hierarchi cal edge orientation matching. Motion adaptive temporal filtering is used to smoothen the pointing trajectories. The algorithm runs in real-time on a standard computer (Pen tium III 700MHz) where it takes only 0.02 sec to process a single frame. We also report first experiences with the algorithm on controlling a virtual mirror application.
Important Links:
Go Back
