Motion capturing from monocular vision by statistical inference based on motion database: Vector field approach

This paper proposes a 3D motion recovery method from monocular images by statistical inference. The fundamental idea of the paper originates from the mimesis model, inspired by the mirror neuron system. The mimesis model is extended to include motion understanding from monocular image sequences and to imitate whole-body motion patterns in 3D space. In order to achieve this goal, (1) conversion of 3D motion database, represented in probabilistic form, into various spaces is adopted. (2) A vector field approach is developed for natural motion understanding. (3) With the particle filter, a demonstrator’s pose is estimated.

I. Introduction

Understanding others from observation is closely related to imagining the self being observed. In such self-imagination, humans convert proprioceptive information to exteroceptive information and vice versa. Here, proprioception is an inte-roception sense of the relative position of neighboring parts of the body and exteroception sense is the perception of the outside world, for example sight.