On Sensor Fusion for Head Tracking in Augmented Reality Applications

The paper presents a simple setup consisting of a camera and an accelerometer located on a head mounted display, and investigates the performance of head tracking for augmented reality applications using this setup. The information from the visual and inertial sensors is fused in an extended Kalman filter (EKF) tracker. The performance of treating accelerometer measurements as control inputs is compared to treating both camera and accelerometer measurements as measurements, i.e., fusing them in the measurement update stage of the EKF simultaneously. It is concluded via simulations that treating accelerometer measurements as control inputs performs practically as good as treating both measurements as measurements, while providing a lower complexity tracker.