Comparative analysis of whole-body center-of-mass estimation methods in dynamic and static activities using marker-based systems

TL;DR

This study compares three methods for estimating the human body's center of mass during static and dynamic activities, finding that whole-body marker sets provide more accurate estimates than pelvis-only methods, especially in dynamic scenarios.

Contribution

It introduces a comparative analysis of CoM estimation methods and evaluates their performance across different activity types, highlighting the effectiveness of whole-body marker sets over pelvis-only approaches.

Findings

Whole-body marker sets outperform pelvis-only methods in dynamic activities.

Incorporating ground reaction forces via Kalman filter does not significantly improve estimates.

Pelvis marker-based method is suitable for static activities but not for dynamic scenarios.

Abstract

Accurate estimation of the whole-body center of mass (CoM) is essential for assessing human stability and postural control. However, selecting the most accurate estimation method is challenging due to the complexity of human movement, diverse nature of activities, and varying availability of equipment, such as marker-based systems and ground reaction force (GRF) sensors. This study compares three CoM estimation methods -- "Pelvis Markerset", "Whole-Body Markerset", and "Whole-Body Markerset & GRFs" -- across static activities, such as standing with eyes closed, and dynamic activities, such as picking up an object from the ground. Using the root mean square (RMS) of "external force residual" (the difference between measured ground reaction forces and estimated CoM accelerations multiplied by total body mass) as a performance metric, we found that while all methods performed similarly under static conditions, the "Pelvis Markerset" method showed 96% to 104% higher RMS external force residual values during dynamic activities compared to the two whole-body methods ($p<0.001$, Cohen's $d$:2.90-3.04). The accuracy of "Whole-Body Markerset & GRFs" (i.e., Kalman filter) was similar to "Whole-Body Markerset", suggesting that incorporating the GRFs through the presented Kalman filter does not improve the estimates from whole-body kinematics. Based on these findings, we recommend the "Whole-Body Markerset" as it performs well and does not require information from GRFs. The "Pelvis Markerset" method can be used in static activities or when markersets around the pelvis reflect whole-body kinematics. This method is not recommended for CoM state estimation in highly dynamic scenarios and when whole-body markersets are available.