Improved accuracy of the whole body Center of Mass position through Kalman filtering

Abstract

The trajectory of the body center of mass (CoM) is critical for evaluating balance. The position of the CoM can be calculated using either kinematic or kinetic methods. Each of these methods has its limitations, and it is difficult to evaluate their accuracy as there is no ground truth to which the CoM trajectory can be compared. In this paper, we use as ground truth the fact that, during the flight phase of running, the acceleration of the CoM is equal to gravity. We evaluate the accuracy of kinematic models of different complexity and find that the error ranges from 14 % to 36 % of gravity. We propose a novel method for optimally combining kinematic and force plate information. When using this proposed method, the error drops to 1.1 to 3.1 % for all kinematic models. Moreover, the proposed method provides a more reliable evaluation of foot placement control, which is commonly used as an indicator of balance during locomotion. The code for calculating this optimal combination is available in both Python and Matlab at: https://github.com/charlotte-lemouel/center_of_mass The documentation is available at: https://center-of-mass.readthedocs.io