Real-time hand gesture recognition by planar and spatial skeletal models

The hand gesture recognition problem is considered. Algorithm to detect planar positions of fingertips by the image of hand silhouette is proposed. It operates with the analysis of continuous skeleton of the hand shape for silhouette segmentation. An extension of the algorithm to spatial case is presented. It uses stereo pair of hand silhouettes to estimate spatial positions of hand and fingertips. The developed methods work in real time, allowing their use in applied hand gesture recognition systems.