Walking Direction Estimation Based on Statistical Modeling of Human Gait Features With Handheld MIMU

Contrary to Global Navigation Satellite System or Wi-Fi based navigation, pedestrian dead reckoning (PDR) method with handheld inertial and magnetic sensors gives the opportunity to achieve indoor/outdoor ubiquitous pedestrian localization. A remaining PDR critical issue is the estimation of the walking direction. Existing methods are principally searching for the energy main axis, but they do not consider the variability of hand movements introducing robustness issues. A new method, based on statistical models and likelihood maximization adjusted to the person and his/her activity, is proposed in this paper. Performance is assessed with experiments in a motion capture room and a shopping mall. The new statistical approach gives globally better results than state of the art methods. A 1.4 degrees to 15.3 degrees error on the walking direction estimates is found over several "1-km walk" tests indoors.