Browsing by Author "Nel, Jason"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Low-Bandwidth transmission of body scan using skeletal animation
    (Stellenbosch : Stellenbosch University, 2020-12) Nel, Jason; Engelbrecht, Herman; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering
    ENGLISH ABSTRACT: With the recent heightened commercial availability of virtual reality head-mounted displays, methods of creating content for such systems are now becoming a relevant concern. Computer generated content is currently widely used for such purposes, as it can easily be geared towards virtual environments viewed on virtual reality displays. Recording content from the real world, however, still poses many challenges. Two common problems in recording video for virtual reality are that the capturing process can require extensive hardware to completely capture a full scene and that large amounts of bandwidth are required to transmit the data generated by 3D video capture hardware. A system is created for this study which attempts to address and avoid those problems, in exchange for a system which is less accurate by means of recreating the experience rather than directly relaying it. As static environments can be captured through other methods, the system records only a human actor. The purpose of this study is to determine whether the animation of a body scan produced by this system can still be an immersive virtual reality experience despite the reduction of accuracy that may be necessary to achieve a low-cost, low bandwidth solution. A single off-the-shelf 3D camera, the Xbox Kinect, is used to capture 3D and skeleton data, as well as other useful data sources such as RGB video and audio. Using only this device allows an easy set-up, at a low cost, while the software approach of animating a body scan with skeleton data allows for low bandwidth transmission over a network. The system created for this study captures, processes, and transmits a body scan over a network in real time. The resultant body scan is then displayed in a virtual environment on a virtual reality head-mounted display. This system achieved a data reduction of over 99% (when compared the original data or a system with similar aims) and the quality was evaluated favourably in a survey of 38 participants recruited to view a demonstration of the system.