OpenSense Calibration

[Mohsen Sharifi Renani]

Hi,

I started playing with OpenSense example and tried to used it on my own dataset, which I collected IMU data from over 20 subjects. I took the following steps but the calibration is not working well for my sensors.
1) performed static trails used for calibrating the model.
2) run dynamic trials and using the calibrated model in the previous step.

Below is the figure after calibration, as you can see the tibia-r imu is not placed correctly. This resulted in an abnormal walking which displayed below. I also noticed abnormal changes in the tibia-r IMU rotations after IMUDataConversion.

To tackle calibration, I tried to rotate the tibia-r IMU and placed it at right location according to the picture of the subject. Then find the rotation transformation between the initial location and modified and used for transfer all tibia-r IMU data to the new location. But, I still got an abnormal trend on my orientation data. What would be the best approach to 1st) improve my calibration model (location of individual sensors) and 2nd) fix the transferred data run by IMUDataConversion.m files.

below is the link for figures.
https://drive.google.com/file/d/17rloEB ... sp=sharing


Data is already collected and I have work with this data.
Thanks,
Mohsen Sharifi
Ph.D. Graduate Student
Center for Orthopaedic Biomechanics, University of Denver

[Giacomo Di Raimondo]

Hi, I was wondering how did you show the reference system of each IMUs in your model on OpenSim? It's really useful. Thanks

[Ayman Habib]

Hi Giacomo,

When you load the calibrated model in the OpenSim application, you can expand the navigator tree for the model to the "imu" object of interest e.g. femur_r_imu (it lives under the "Bodies/femur_r" node), then you can use the context menu (right-mouse click on windows) to "Show Axes". The axes are color coded RGB for XYZ.

Hope this helps,
-Ayman

[Giacomo Di Raimondo]

Hi Habib

In did the calibration procedure with Matlab, in my case I have the -z axis as the forward one, but when I defined (-pi/2, 0, 0) as my rotation, what I get in the calibrated model is different (see picture attached).
Is this because the -z axes is not perfectly aligned with x?
How do I know the rotation I need? It seems as if the calibration pose of the subject in a different direction?

https://kuleuven.box.com/s/dzuz9tgukbrq ... 6c9agcgbpu

[Ayman Habib]

Hello,

Please send image/capture as the link you sent is broken.

Thank you,
-Ayman

[Giacomo Di Raimondo]

IMU_on_model.PNG (313.4 KiB) Viewed 335 times