Model-based 3D tracking of an articulated hand from single depth images

Abstract

This paper presents a novel solution to the problem of tracking the 3D position, orientation and full articulation of a human hand from single depth images. We choose the model-based approach and treat the tracking task as an optimization problem. A new objective function based on depth information is presented to quantify the discrepancy between the appearance of hypothesized instances of a hand model and actual hand observations. Sequential Particle Swarm Optimization method is proposed to minimize the objective function for sequential optimization. An semi-automatic hand location method is adopted to predict hand region for sequential tracking. A GPU-based implementation of the proposed method is well designed to address the computational intensity. Extensive experimental results demonstrate qualitatively and quantitatively that tracking of an articulated hand can be achieved in real-time.