Page 1 of 1
Suspicious GRF data
Posted: Sun Apr 21, 2024 11:40 am
by alexdenton
Hello,
I am including GRF data along with my moco cap data in my AddBiomechanics process. I noticed my log file includes:
Detected missing/bad GRF data on 54/54 frames
ERROR: we have no good frames of GRF data left after filtering out suspicious GRF frames. This probably means input GRF data is badly miscalibrated with respect to marker data (maybe they are in different coordinate frames?), or there are unmeasured external forces acting on your subject.
The GRF data look sufficient in the OpenSim GUI. Any suggestions?
Thanks!
Alex
Re: Suspicious GRF data
Posted: Tue Apr 23, 2024 1:57 pm
by nbianco
Hi Alex,
I would try uploading only a model and marker data to AddBiomechanics to run a kinematics-only analysis to get joint angles. Then, use those joint angles and the GRF data to compute inverse dynamics in OpenSim. That way you can evaluate the quality of the GRF data based on the resulting joint moments.
If the joint moments look right, but AddBiomechanics still complains, then we can investigate deeper to see if there is a bug.
-Nick
Re: Suspicious GRF data
Posted: Wed Apr 24, 2024 1:30 pm
by alexdenton
Hi Nick,
Thank you so much for the feedback. I ran the kinematics-only process in AddBiomechanics, followed by ID in the OpenSim GUI with GRF data. The joint moments were not great.
When processing one full gait cycle in AddBiomechanics the provided GRF are not disregarded as suspicious, though the force and torque residuals are not great. When processing either one left or right step, the GRF data are disregarded and marked as suspicious.
Alex
Re: Suspicious GRF data
Posted: Wed Apr 24, 2024 3:39 pm
by nbianco
Hi Alex,
When processing only one step, are you still uploading GRF data for both feet? At each time stpe, if AddBiomechanics can't match a GRF source to a foot body in the model it will mark the GRF data as "suspicious".
-Nick
Re: Suspicious GRF data
Posted: Thu Apr 25, 2024 9:20 am
by alexdenton
Hi Nick,
Yes, I am uploading GRF for both feet when processing just a single step. It's a typical .mot with 19 columns: 1 for time and 9 for each side. In other words, when processing a left step in AddBiomechanics, I am uploading a .mot with data in columns 2-10, and columns 11-19 are filled with 0s. It loads into the GUI just and visually looks as expected.
I just tried processing a step with GRF data (.mot contained 10 columns) for only the contacting foot but it was still marked as suspicious.
Alex