Real-time and precise 3-D hand posture estimation based on classification tree trained with variations of appearances

We propose a method for estimating 3-D hand postures from 2-D monocular images in real-time. The estimation is based on finding the best matched posture from typical postures whose appearances are learned in advance. For high accuracy, conventional methods require high computational cost for comparing an input with many typical postures. In our method, a tree is automatically generated and trained with typical postures and their variations. Efficient search with the tree brings about real-time estimation. We show the effectiveness of our method by some experimental results.