Point clouds indexing in real time motion capture

Abstract

Today's human-computer interaction techniques are often gesture-inspired and thus pushed towards naturalness and immediateness. Their implementation requires non-invasive tracking systems which can work with little or no body attached devices, like wireless optical motion capture. These technologies present a recurrent problem, which is to keep a coherent indexing for the different captured points during real time tracking. The inability to constantly distinguish tracked points limits interaction naturalness and design possibilities. In this paper we present a real time algorithm capable of dealing with points indexing matter. Compared to other solutions, the presented research adds a computed indexing correction to keep coherent indexing throughout the tracking session. The correction is applied automatically by the system, whenever a specific configuration is detected. Our solution works with an arbitrary number of points and it was primarily designed for fingertips tracking. A Virtual Reality application was developed in order to exploit the algorithm functionalities while testing its behavior and effectiveness. The application provides a virtual stereoscopic, user-centric environment in which the user can trigger simple interactions by reaching virtual objects with his/her fingertips.