6DoF Tracking in Virtual Reality by Deep RNN Model

Abstract

In this paper, a novel coordinates tracking method is proposed for Virtual Reality (VR) environment using sensor signals. The purpose is to extend movement tracking in VR from 3 Degrees of Freedom (DOF) of rotation to 6DOF of position plus rotation. As a result, we can track VR coordinates without using controller or handler provided by VR devices. An RNN-based model is proposed to predict displacement of positions in each timestamp given measured acceleration and Euler angles from sensor signals. Experiments demonstrate that it is effective to predict correct position displacement, which not only models the relationship between sensor signals and displacement but also handles the cumulative errors during tracking.