User Model for Predictive Calibration Control on Interactive Screens

On interactive surfaces, a precise calibration of the tracking system is necessary for an exact user interaction. So far, common calibration techniques focus on eliminating geometric distortions. This static calibration is only correct for a specific viewpoint of one single user and parallax error distortions still occur if this changes (i.e. if the user moves in front of the digital screen). In order to overcome this problem, we present an empirical model of the user's position and movement in front of a digital screen. With this, a model predictive controller is able to correct the parallax error for future positions of the user. We deduce the model's parameters from a user study on a large interactive whiteboard, in which we measured the 3D position of the user's viewpoint during common interaction tasks.