A Procedure for Adjustment of Body Segmental Parameter Values to Individual Subjects in Inverse Dynamics

Akinori Nagano, Taku Komura, Ryutaro Himeno, Senshi FukashiroInternational Journal of Sport and Health Science 2004 · DOI

Abstract

A procedure of inverse dynamics was developed to adjust the body segmental parameter values to individual subjects. Newton's second law was utilized, which states that a resultant force vector (the sum of all forces acting on a rigid body) can be calculated from the mass of the body and the acceleration vector of the center of mass of the body. By comparing the measured resultant force and the calculated resultant force, it was possible to evaluate the errors that exist in body segmental parameter values. These errors were minimized through simulated annealing numerical optimization that searched for the optimal values of body segmental parameters. A three-dimensional neuromusculoskeletal model was used to generate error-free sets of kinematic and kinetic data. Two types of jumping motion, i.e., squat jumping and countermovement jumping, were generated through forward dynamic computer simulation. In the process of analysis, randomly generated errors were introduced into body segmental parameter values, i.e., the mass and the location of the center of mass of each segment. The procedure developed in this study successfully reduced the errors in those body segmental parameter values. The average error for the mass was 0.97% whereas the average error for the location of the center of mass was 6.04%.

Introduction

In biomechanics, inverse dynamics is the standard method for calculating internal joint forces and moments during movement. However, the accuracy of these calculations relies heavily on knowing the precise mass, length, and center of mass (CoM) for every body segment. Traditionally, researchers have relied on population-average anthropometric tables. This approach introduces significant errors when a subject deviates from the "average" body type. This paper proposes a numerical optimization procedure to move away from generic averages and toward subject-specific models using only standard motion capture and force plate data.

Methodology

The core of the calculation relies on Newton’s Second Law ($F = ma$), where the sum of the mass-acceleration products of all segments must equal the total ground reaction force (GRF) minus gravity. The researchers utilized a "synthetic data" approach to validate their optimization routine:

  1. The Model: A 3D musculoskeletal model was built with nine rigid segments and 20 degrees of freedom.
  2. Generating "Ground Truth": Forward dynamic simulations generated "perfect" kinematic and kinetic data for two movements: a squat jump and a countermovement jump.
  3. Introducing Error: Segment parameters (mass and CoM location) were intentionally corrupted with random errors ranging from 50% to 150% of their true values.
  4. The Optimization: Simulated Annealing (SA)—a global optimization algorithm—was used to search for parameter values that minimized the discrepancy between calculated and measured GRFs.
Figure 1: Flowchart of the numerical optimization procedure using simulated annealing to adjust body segmental parameters.

Key Results

The procedure was highly successful in reducing the intentionally introduced errors. The Root Mean Square (RMS) difference between the calculated and measured ground reaction forces dropped from 403.382 N (before optimization) to just 2.139 N (after optimization).

Figure 2: Ground reaction force (GRF) profiles for squat jumping (A) and countermovement jumping (B) after optimization, showing minimal RMS error.

Specific accuracy metrics included:

  • Mass Accuracy: The average error in segment mass was reduced to a mere 0.97%.
  • CoM Accuracy: The average error for the location of the center of mass was reduced to 6.04%.

Sensitivity analysis revealed that the optimization was significantly more accurate for larger segments (like the trunk) than for smaller ones (like the toes). This is because heavy segments contribute much more to the total ground reaction force signal; errors in smaller segments are more easily lost in the noise.

Discussion and Conclusion

This study demonstrates that subject-specific tuning of body parameters is possible without specialized medical imaging like DEXA or MRI. By using standard laboratory equipment, researchers can significantly improve the accuracy of inverse dynamics.

A primary limitation is the use of "clean" computer-generated data. In real-world settings, skin movement artifacts and sensor noise would add complexity. Future work should focus on testing this methodology on "non-standard" populations—such as children, the elderly, or athletes with extreme body proportions—where standard anthropometric tables are known to fail.

Summary updated

Abstract

A procedure of inverse dynamics was developed to adjust the body segmental parameter values to individual subjects. Newton's second law was utilized, which states that a resultant force vector (the sum of all forces acting on a rigid body) can be calculated from the mass of the body and the acceleration vector of the center of mass of the body. By comparing the measured resultant force and the calculated resultant force, it was possible to evaluate the errors that exist in body segmental parameter values. These errors were minimized through simulated annealing numerical optimization that searched for the optimal values of body segmental parameters. A three-dimensional neuromusculoskeletal model was used to generate error-free sets of kinematic and kinetic data. Two types of jumping motion, i.e., squat jumping and countermovement jumping, were generated through forward dynamic computer simulation. In the process of analysis, randomly generated errors were introduced into body segmental parameter values, i.e., the mass and the location of the center of mass of each segment. The procedure developed in this study successfully reduced the errors in those body segmental parameter values. The average error for the mass was 0.97% whereas the average error for the location of the center of mass was 6.04%.
BibTeX
@article{nagano2004procedure,
  author = {Akinori Nagano and Taku Komura and Ryutaro Himeno and Senshi Fukashiro},
  title = {{A Procedure for Adjustment of Body Segmental Parameter Values to Individual Subjects in Inverse Dynamics}},
  journal = {International Journal of Sport and Health Science},
  year = {2004},
  doi = {10.5432/ijshs.2.156},
}
Figure