Authors:
Fernando Losilla
and
Francisca Rosique
Affiliation:
Universidad Politécnica de Cartagena and Spain
Keyword(s):
Exergame, Augmented Reality, Body Pose Estimation, OpenPose.
Related
Ontology
Subjects/Areas/Topics:
Informatics in Control, Automation and Robotics
;
Robotics and Automation
;
Virtual Environment, Virtual and Augmented Reality
Abstract:
Exergames have become very popular for fitness and rehabilitation purposes. They usually rely on RGB-D sensors to estimate human 3D body pose and, therefore, allow users to interact with virtual environments. Currently, a new generation of deep learning techniques enable the estimation of 2D body pose from video sequences. These video sequences could be augmented with the estimated pose and other virtual objects, resulting in augmented reality mirrors where players can see their reflection along with other visual cues that guide them through exercises. The main benefit of using this approach would be replacing RGB-D cameras with simpler and more widely available webcams. This approach is explored in this work with the development of the ExerCam exergame. This application relies on a single webcam and the OpenPose library to allow users to perform exercises where they have to reach virtual targets appearing on the screen. A preliminary study has been performed in order to explore the
technical viability and usability of this application, with promising results.
(More)