About the implmetation of RBF

[NgChiho]

Dear Mr Lund:

Hello! Sorry for disturbing you in such a sudden.

Lund-2015-Scaling-of-musculoskeletal-models-f.pdf

I have read an article of yours, which is "Scaling of musculoskeletal models from static and dynamic trials"(Internetional Biomechanics, 2015), and been impressed by the RBF method used to implement the non-linear scaling of a model. However, I don't quite understand the concrete implmentation process of this method, so I want to ask some questions and hope for your answers, which are:

1. Did you only use the markers of motion capture data to generate the scaled model? Or did you combine with some points of the origin model?
   (Because I think only by the markers it may not be sufficient to generate a complete model.)
2. May I know more about the RBF calculation process during scaling, like how to calculate the coefficients of the radial basis function?
3. Have you tried this method in upper body? If yes, can it lead to a satisfactory result?

[melund]

@NgChiho

Checkout the supplementary material for the paper. It gives bit more information than what is in the paper:

https://www.tandfonline.com/doi/full/10.1080/23335432.2014.993706

Did you only use the markers of motion capture data to generate the scaled model? Or did you combine with some points of the origin model? (Because I think only by the markers it may not be sufficient to generate a complete model.)

Well I am not completely sure what you are asking here. The kinematic part/mechanism (stick figure) of the model was generated solely from the standing reference markers.

The full musculoskeletal model was then morphed to match the stick-figure model. So the full is of course not generated completely from the markers. But it is the markers that solely define the kinematic structure of the model.

May I know more about the RBF calculation process during scaling, like how to calculate the coefficients of the radial basis function?

I believe there is bit more information in the supplementary material on RBF. However, I didn't implement the RBF functions from scratch. Instead I use the implementation provided with the AnyBody Modeling System. It allows me to easily create and RBF interpolation function between two sets of source and target points. I did however check that the results were the same as the RBF package which comes with the Scipy python package: https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.Rbf.html

Have you tried this method in upper body? If yes, can it lead to a satisfactory result?

No, unfortunately not. I would very much have liked to investigate this, but I never got the chance. I think it would be more complicated as the arms and shoulder joint have more degrees of freedom. Also, there isn't such a well established marker protocols for the upper body to base the work on.

Best wishes,
Morten