2D-to-3D Visual Human Motion Converting System for Home Optical Motion Capture Tool and 3-D Smart TV

Although 2-D visual human motion analyzers that work with a sequence of images have been widely studied, some consumer electronics equipment, such as home optical motion capture tools and smart TVs, require 3-D motion data. For this purpose, this paper proposes a 2D-to-3D visual human motion converter that reconstructs 3-D visual motion from the 2-D visual human motion in a sequence of images. Because home optical motion capture tools and smart TVs have a single channel, this paper designs a proposed 2D-to-3D converter that has a single channel. A common difficulty for the single channel is that most contemporary 3-D reconstruction processes based on 2-D motion are nonlinear iterative types. For more convenient use in real-time applications, this paper proposes an analytical solution type of 3-D reconstruction process instead of the nonlinear iterative types. To derive an analytical solution, this paper proposes an idea that formulates a human body joint as a 2-D universal joint model instead of the more common 3-D spherical joint models. To overcome the major limitation of analytical solution-type processes, i.e., the low accuracy, this paper formulates the estimation process as an optimization problem. The process thus designed is then applied to each joint in the human body, one after another. Performance evaluation shows that the proposed 2D-to-3D visual human motion converter has nearly the same high accuracy as a contemporary nonlinear iterative-type approach.1.