Gesture Variation Estimation for Whole-Body Movement Interactions

Abstract

Abstract Department of Computer Science Bachelor of Arts Gesture Variation Estimation for Whole-Body Movement Interactions by Alejandro Van Zandt-Escobar This paper describes the motivation, design and implementation of a system for real-time whole-body gesture analysis, intended for interactive performance applications exploring the relationship between movement and sound. We describe a template-based recognition method called meta-Gesture Variation Followers (mGVF), which builds upon the Gesture Variation Follower, extending it from single-point input to accept a set of spatial coordinates representing the human body skeleton, which are treated with a hierarchical approach which takes into account the underlying structure. The system performs online gesture recognition and variation estimation. Finally, we present an application in which the system is used for an interactive sonic performance in which a user's movement is used to control a sound synthesis engine.