Uncertain input estimation with application to Kalman tracking

Abstract

Many motion tracking systems average and integrate tracking measurements over a period of time in order to reduce the effects of device noise, external noise and other disturbances. The target (user) is likely to be moving throughout the sample time, introducing additional 'noise' (uncertainty) into the measurements. Without filtering, noise can cause small variations in the estimated tracking positions (tracking drift) over time. There are many filters and algorithms that account for uncertainty due to noise. The Kalman filter has been chosen in this study because of its ability to estimate tracking positions and to account for uncertainty in the tracked object's position where it is occluded by other stationary or moving objects. An inexpensive algorithm is presented which detects the slightest motion and then tracks the motion or the target very accurately.