Inertial pedestrian localization with soft constraints based on biomechanical models

Abstract

To this day, inertial localization systems are failing to incorporate biomechanical constraints in the position tracking process. The goal of this paper is to define biomechanical constraints on the attitude estimates of an inertial localization system. To that end, a biomechanical study of the human leg is carried out in order to model the attitude as a set of Gaussian distributions. The latter impose soft constraints on the attitude that can be optimally integrated in a Kalman filter. Therefore, the constraints are integrated in the respective unscented Kalman filters (UKF) of two inertial navigation systems (INS): a thigh-mounted INS and a foot-mounted INS. The effect of these constraints in the position estimation is evaluated with a dataset of, approximately, 5h. The results show that the biomechanical constraints improve the coherence of the roll and pitch estimated by the thigh-mounted localization system. The impact of these constraints is reflected in the height error of the thigh-mounted system, which is improved by, approximately, 82%. The biomechanical constraints make the roll and pitch of an inertial localization system coherent with respect to human motion. Nevertheless, the constraints do not improve the heading estimated by the localization systems.