This paper describes a method for generating 3D computer graphics animation of hand-gesture using a parameter generation algorithm based on hidden Markov model (HMM). The purpose of this study is to generate motion of gestures represented by a label sequence. Each label represents a basic motion pattern of the hand, which is modeled by an HMM. When modeling a basic motion pattern by HMM, gesture data, which are parameter sequences of a physical model of the hand, recorded using motion-capturing are used as training samples. Then, given a label sequence, an HMM is composed by concatenating HMMs in the order according to the label sequence, and then a gesture is generated from the composed HMM in a maximum-likelihood sense and put into a computer graphics animation. Smoothness of the synthetic gesture can be achieved by using statistics of static and dynamic features modeled by HMMs. An experimental result shows the effectiveness of this synthesis method.
展开▼
