Invited talk: Copycat hand for all at Laval Virtual ReVolution 2009 (ヒューマン情報処理)

概要

論文の詳細を見る
We propose a hand posture estimation system that can search a hand posture image most similar to the input data from a large-scaled database. Our system uses, at the first stage, coarse screening by the proportional information on the hand images which roughly correspond to forearm rotation or bending of the thumb or four fingers. And then, at the second stage, performs a detailed search for similarity for selected candidates. To describe forearm rotation, and wrist's internal and external rotation, bending and stretching, no separate processes were used for estimating the corresponding joint angles. We have also enhanced flexibility for hand size and form, e. g. those having thicker fingers, longer fingers, or deeply hooking fingers, through using different typical fingers CG images featuring different bone length and thickness and joint movable ranges in order to permit the system to respond accurately to people of all ages and both sexes, including foreign people. By estimating the sequential images of the finger shape by using this method, we successfully realized a process involving a joint angle estimation error within several degrees, a processing time of approximately 80fps or more, by using only one Note PC and high-speed camera, even when the wrist was freely rotating.
2009-06-08