Beyond Mocap: Animating Soccer Players Based on Positional Tracking Data

Abstract

Due to current advances made in Virtual Reality solutions, we can now simulate real world professional soccer matches in such a way that anyone can relive a match as if they were on the pitch. Such simulation can be built through the use of motion capture clips and positional tracking data captured during a match to replicate the player’s motions on the virtual field. However, the automation of motion capture usage is limited due to the lack of pose information. We propose a framework, Beyond Mocap, which takes these two data sources and, together with an annotation system, automatically generates the desired motion. Beyond Mocap works by taking user provided descriptions of the moves performed by the soccer player and matching them with an annotated motion dataset. To ensure the virtual character follows the same path and with minimal foot-skating, we present a novel approach of travelled distance matching by taking advantage of the fact that locomotion is cyclical and, as such, it can be made longer or shorter by adding or removing locomotion cycles. A simple method of motion dataset duplication by mirroring is also provided, enabling a wider range of available motion capture. Beyond Mocap outputs motions that are more realistic-looking, have a higher degree of naturalness and are smoother than the motions produced by the current solution, while at the same time following the path defined by the positional tracking data and with minimal foot-skate.