Error in kinematic analysis

[endresd (Dominik Endres)]

Hello everyone,

seventeen subjects have been recorded by Xsens MVN Awinda inertial motion capture system with 17 wireless motion trackers and an update rate set to 60 Hz.
I am interested in estimating the Ground Reaction Force out of the BVH files. Here, the Xsens T-pose is in the beginning of every BVH file I am going to analyse. Thus, I was loading the inverse dynamic model, running the BVH_PreProcess and started to run the analysis. Some estimations succeeded, others resulted in errors. While analysing the BVH data of two different subjects I got errors during kinematic analysis. It is said that the constraints are above tolerance.
Please find one screenshot attached. I would be happy if you could assist me in how to proceed.

Many thanks and best regards,
Anja

[toerholm (Søren Tørholm)]

Hi Anja,

Sorry for the slow response

Please try to avoid having the T-pose in the export since this may cause problems, when the model needs to go from t-pose to real position within one frame.

So either manually delete the frame in the bvh file or export the bvh again.

Hope it helps.

Best regards
Søren

[endresd (Dominik Endres)]

Hi Soren,

Thanks for your reply. In my data the T-pose is captured within the first frame. Thus, I adjusted the StartFrame in trialspecificdata.any. I set the StartFrame to ‘2’, loaded the model, started the BVH_PreProcess and ran the analysis, but I still got an error in kinematic analysis. Then, I set the StartFrame to ‘7’ to be absolutely sure that there is no T-pose related data in the future analysis. However, again I got an error in kinematic analysis.

I am happy about any suggestions.

Thanks,
Anja

[CDzialo (Christine Dzialo)]

Hi Anja,

Can you attach the bvh file here, along with the subjects height and weight please so we can see what the error is?

Best Regards,
Christine

[endresd (Dominik Endres)]

Hi Christine,

please find the bvh file with corresponding body dimensions in the below folder.

Thanks for your help,
Anja

VP2-002_kinematic_error.zip (4.6 MB)

[Lewis-H (Lewis Hamilton)]

Thanks this information helped! But this is kind of long process. Are there any shortcuts?
Regards,
Lewis

[CDzialo (Christine Dzialo)]

@Lewis-H Please create a new topic for this and state the specific question(s) you have.

Best Regards,
Christine

[CDzialo (Christine Dzialo)]

Hi Anja,

I am running only a portion of the trial right now and not obtaining any errors. What step did your model fail?

Best Regards,
Christine

[endresd (Dominik Endres)]

Hi Christine,

when I am running the model with StartFrame setting to 7, the analysis fails in time step 12857.

Best regards,
Anja

[CDzialo (Christine Dzialo)]

Hi Anja,

Can you try this solution:

Go into BVHSettings.any and switch Rotation1PiFixOnOff = Off;
Rerun the BVH_Preprocess and then try again to RunAnalysis.

If it still doesn't work, try switching Rotation2PiFixOnOff= Off; as well and rerun BVH PreProcess.

Please see this post for more information:

AnyMoCap+Xsens - Error in Euler angle discontinuity filtering

Hi Marco, The problem appears to be that there are angular jumps in the BVH data. The AnyInputBVH object has filters which can be used to remove these jumps of e.g. 2pi or 1pi see the Rotation2PiFixOnOff or Rotation1PiFixOnOff objects. The error you get is because AnyBody is trying to apply this filter but while doing this it detects that the jumps on different rotations in the joint do not match perfectly and throws an error. I tried to load your file and disable the angular filtering of ju…

Best Regards,
Christine

[CDzialo (Christine Dzialo)]

Hi Anja,

I wanted to point your attention to this new post that may help fix the problem you are having (rather than using the work-around I gave previously):

AnyBody Modeling System v.7.2.3 relesed

We are pleased to announce that the AnyBody Modeling System v.7.2.3 and AMMR 2.2.3 has just been released. It can be downloaded from the AnyBody Technology download page. AMS v.7.2.3 is a bug fix release which mainly targets problems with models using BVH files, but also fixed a number of other smaller issues: Minor updates AnyInputBVH: A new mode improving the options to make data interface to the human model comprising: Better options to deal with discontinuous (jumping) angular data…

Here is another document on how to update your old model:

Hope this helps, have a good weekend

Best Regards,
Christine

[endresd (Dominik Endres)]

Hi Christine,

both approaches you have recommended above worked, thanks a lot.
However, in the updated environment (approach 2, v.7.2.3, AMMR 2.2.3) the inverse dynamic analysis took about 6 days whereas the first approach (Rotation1PiFixOnOff = Off with AMMR 2.2.1) took only 1 day. For time reasons I would prefer approach 1. Can you tell me what Rotation1PiFixOnOff = Off causes? Do you think it is harmful for acurate data?

Thanks,
Anja

[toerholm (Søren Tørholm)]

Hi Anja,

6 days sounds as a lot, how many frames do you have in the study?

Normally it can be a good idea to split long analysis into several to avoid having too many frames in the study.

The Rotation1PiFixOnOff is a setting that allows a filter to be enabled which will look at the rotation data in the BVH file and seek to remove jumps of 1PI. If this filter is off it will not be applied so if the data has these jumps the kinematic analysis may stop. If you do not see kinematic problems with the setting to be off you will usually be fine. If the filter is on it can sometime cause problems because the jumps in the angles may not always be exactly 1PI.

The new version 7.2.3 should resolve all issues with jumps in angles, we have not noticed a slow down but we will check it, is this something you see consistently also for models with fewer frames?

Best regards
Søren