Toward automatic robot programming: Learning human skill from visual data

We propose a no,el approach to program a robot by demonstrating the task mu ltiple number of times in front of a binocular vision system. We track arti ficially-induced features appearing in the image plane due to nonimpediment al color stickers attached at different fingertips and wrist joint, in a si multaneous feature detection and tracking framework. A Kalman filter does t he tracking by recursively predicting the tentative feature location and a higher order statistics (HOS)-based data clustering algorithm extracts the feature. A fast and efficient algorithm for the vision system thus develope d processes a binocular video sequence to obtain the trajectories and the o rientation information of the end effector from the images of a human hand. Thf concept of trajectory bundle Is introduced to avoid singularities and to obtain an optimal path.