Dear AnyBody team,
I have some issues with the marker tracking. Please see the picture attached below.
We can see the sentence "Kinematic analysis failed in time step 37：Position analysis is not completed" displayed at the bottom of the picture.
I have two questions。
First，I have checked the distance between the model points and the motion capture points, which are almost identical, but the kinematic analysis still cannot be completed. How can I locate the problem I encountered during my exercise studies and modify it through step 37?
The second question is, why do some of the coordinate axes of AnyBody's model points in the picture are green (number 1 in the picture) and some are red (number 2 in the picture)? What are the meanings represented by these two colors?
Please help me explain so that I can modify the program to complete kinematic analysis.
Your feedback will be highly appreciated. Thanks
I would suggest you go through our tutorial on working with marker-based data at this link. That will explain you about the red and green colored coordinate systems.
Once you have gone through the tutorial,
Then, my question for you: have you run the ParameterIdentification process?
Also, from the image, the left talus looks in a strange position. Can you please check if the markers are correctly identified? It's possible that the markers are incorrectly identified in the AnyBody model or in the C3D data. Please check both to be sure.
I hope this helps.
I have a similar problem with my kinematic analysis. Hope you managed yours Haod.
I also see the warning: "Kinematic analysis failed in time step 41：Position analysis is not completed". But unfortunately I can´t run the parameter identification process. There is only an option called "Studies.ParameterIdentification.Paramterer Optimisation", which doesn´t help me.
Also I am using BVH data instead of C3D data. I don´t know if this is part of the problem.
Maybe you solved similar problems yet ?
Thanks for your help and kind regards,
I am sorry to read that. Is it possible for you to share a screenshot of the model view when your analysis fails? Do you see some awkward posture? Also, can you please share some details about your data collection: which system did you use to record the motion? What frequency? Is it possible that you have the T-Pose or N-Pose as the first frame in your bvh file?
Regarding your analysis using bvh data, there is no parameter optimization process for the standard bvh pipeline. In the bvh models, the markers are added virtually to the stick figure generated from the bvh data and the human model. Also, the segment dimensions are directly used from the bvh data. The parameter optimization option exists in case some models need it (for e.g., Optimize BVH Origin example — AMMR v3.0.0 Documentation).
thank you for your help.
Here is a screenshot of my Model View.
I am quite sure there is no strange posture in any of my failed inverse dynamics studies as far as I can see.
I used the XSens Motion Capture (Movella) system with 17 full body sensors and 60Hz frequency. And yes normally I start the movements of the patients in the N-Pose (not striktly or by purpose, but they just normally stand as they like). Could that be a problem? Because then the solution should be to change the first frame in the 'TrialspecificData'. But if I do so the problem sill remains.
Some of my measurements just run normally in the anybody program and some get this error message.
Thanks for the help and information.
With Xsens, you have the option of appending the first frame as T-Pose. This doesn't work very well as there is then a big jump between the first and second frame. In my opinion, the best practice is to start the recording with the subject in the N-pose (standing relaxed) and then the subject performs the motion. From what you wrote, I think you have done exactly that: subject standing in the N-Pose and then doing the motion. Also, you are right that if you change the first frame from TrialSpecificData.any, you would be able to avoid this problem altogether.
One other question about Xsens: did you HD process your motion data before exporting to bvh? If yes, then I would like to try to run the bvh file myself and investigate a little bit. I can't really think of another reason at the moment. Would it be possible for you to share the bvh file? If it's a big bvh file, please crop it to around 200 frames around the frame where the analysis is failing (or alternatively, please let me know the failing frame number and I can adjust that in TrialSpecificData). Also, if you share the bhv file, can you please share the subject height and weight?
One more thing I noticed: Is there any specific reason you have enabled Shoulder Rhythm? Can you try to disable it and run the analysis again? Shoulder rhythm will probably fight with the marker data so this may lead to some errors.
thank you for the informations and your help. Actually my analysis is running when disabling the Shoulder Rhythm! So maybe the marker data was really fighting the shoulder rhythm. Thanks for the solution!
I enabled the Shoulder Rhythm because I thought this way I get the most realistic movement and it was also written in the example configuration of the Shoulder Arm Model. But if I understand it correctly, my own marker data is even more realistic than the strict shoulder rhythm one can enable in the script? So there is no disadvantage for disabling it for me, because my real marker data is even better?
Thank you and kind regards,
You are right about shoulder rhythm. Its purpose is to simulate realistic shoulder complex movement by just specifying the glenohumeral joint angles. This is for cases when mocap data is not available.
However, when you have your own motion data, the mocap data will typically prescribe the movement at the shoulder complex (e.g, marker placed on shoulder). There is no need to include shoulder rhythm then. As you saw, shoulder rhythm can create conflicts with the marker data and result in failure of marker tracking.
Ah okay perfect! Thank you very much for your help.