Real-Time 3D Hand Interaction: Optimization and Complexity Reduction

Paper in proceeding, 2008

This paper suggests a low-cost way to enable 3D hand interaction based on the frame capturing functionality of a single consumer-grade webcam. Our recent developments have focused on examining and realizing a less complex method. The suggested method reduces tracking effort to only one reference marker: a color-coded bracelet that helps locate the part of the captured frame containing the user's hand. The located area contains all the information needed to extract hand rotation and finger angle data. To facilitate hand feature extraction we outfit the user's hand with a specially coded glove equipped with two square palm markers, one for each side of the hand, and five distinct finger colors. We believe that our approach of tracking only one marker will be more efficient than related methods that track each finger separately. The method is further simplified by using spatial properties drawn from physiological characteristics of the human hand to limit the areas considered by the algorithm. Some challenges in realizing this method are related to webcam limitations such as image noise and limited image- and color-resolution. Other challenges are overlapping hands and fingers, hand positioning outside the field of view, and interference by local light sources.