Abstract

The comparison and implementation of two different approaches for the automatic detection and recognition of varying body contacts in MUVR are at the heart of this thesis. This is done to find a best practice implementation when large databases to train machine learning algorithms are lacking. After an analysis of the current state of the art in pose and gesture recognition and identification of the unique advantages that VR offers for this, two approaches (Collider-based vs. Vector-based recognition) are presented in detail. These are then implemented for a couple of exemplary body contacts of varying difficulty and tested at certain milestones to inform the development of the next milestone.