Hallo everybody,
I try to implement IMU data by Perception Neuron in OpenSim through OpenSense. But it's not working, in the following link you can find a video in which I show my problem.
https://drive.google.com/drive/folders/ ... sp=sharing
The coordinate systems should correspond as AxisNeuronPro orientation is North(x)-Up(y)-East(z).
I appreciate any idea about what could be the reason for my problem.
Cheers
Johannes
OpenSense implementation problems
-
Johannes Hänlein
- Posts: 4
- Joined: Thu Mar 05, 2020 1:14 am
-
Ayman Habib
- Posts: 2243
- Joined: Fri Apr 01, 2005 12:24 pm
Re: OpenSense implementation problems
Thanks for the nice reporting.
There're many questions however that are unanswered by your report:
1. What kind of data is reported by the imu system or test data (we exclusively use orientations, and assume sensor fusion was done before hand). You mentioned reading bvh files but not what data is included there.
2. Did you perform calibration as suggested by the OpenSense documentation? If so there are many other questions you need to answer/resolve:
2.a Did you apply a transform from your IMU system to OpenSim?
2.b Did you perform heading correction?
2.c Did you visualize the calibrated model and see that the IMUs are placed in the expected orientation on corresponding segments as the test data (that's accessible to you)
We're in the process of developing visualization tools to help with troubleshooting, it is still work in progress but should become available over the next couple of months, but hope that by answering the above questions you can pinpoint problems/issues.
Hope this helps,
-Ayman
There're many questions however that are unanswered by your report:
1. What kind of data is reported by the imu system or test data (we exclusively use orientations, and assume sensor fusion was done before hand). You mentioned reading bvh files but not what data is included there.
2. Did you perform calibration as suggested by the OpenSense documentation? If so there are many other questions you need to answer/resolve:
2.a Did you apply a transform from your IMU system to OpenSim?
2.b Did you perform heading correction?
2.c Did you visualize the calibrated model and see that the IMUs are placed in the expected orientation on corresponding segments as the test data (that's accessible to you)
We're in the process of developing visualization tools to help with troubleshooting, it is still work in progress but should become available over the next couple of months, but hope that by answering the above questions you can pinpoint problems/issues.
Hope this helps,
-Ayman
-
Johannes Hänlein
- Posts: 4
- Joined: Thu Mar 05, 2020 1:14 am
Re: OpenSense implementation problems
Hi Ayman,
Thank you for your help!
1. Data type is a calculation file .calc, I export this from Axis Neuron Pro (https://neuronmocap.com/content/axis-neuron-pro) it's a sample data and I only use the orientations in quaternion. I assume strongly that Perception Neuron is using sensor fusing (it's a comercial motion capture system), I don't use .bvh files, because I can't find the needed data for OpenSense in it.
2. I followed the OpenSense documentation after I transformed the .calc file to .csv, therefore I simply run the calc2csv.m script in Matlab (it's scripted by myself)
2.a I transformed the coordinate system by a rotation of pi/2 around the x-axis, because I believe in Axis Neuron Pro z-axis is down, x-axis to the left.
2.b I tried a heading correction, but I am not sure whether I need one or in which direction.
2.c As I said above this is only a sample data and I can't see the original orientation in Axis Neuron Pro, but they are at the correct segment.
I added a new folder ("17N in Tristan's Office") with all the mentioned data to my drive. https://drive.google.com/drive/folders/ ... sp=sharing, after running the calc2csv.m file you just need to run the apdm_test.m file. my current transformation and heading correction you can find in the .xml files.
Could the problem occure, because the subject in the calibration data doesn't exactly match the default pose of the model?
I would appreciate any further help to this topic!
regards
Johannes
Thank you for your help!
1. Data type is a calculation file .calc, I export this from Axis Neuron Pro (https://neuronmocap.com/content/axis-neuron-pro) it's a sample data and I only use the orientations in quaternion. I assume strongly that Perception Neuron is using sensor fusing (it's a comercial motion capture system), I don't use .bvh files, because I can't find the needed data for OpenSense in it.
2. I followed the OpenSense documentation after I transformed the .calc file to .csv, therefore I simply run the calc2csv.m script in Matlab (it's scripted by myself)
2.a I transformed the coordinate system by a rotation of pi/2 around the x-axis, because I believe in Axis Neuron Pro z-axis is down, x-axis to the left.
2.b I tried a heading correction, but I am not sure whether I need one or in which direction.
2.c As I said above this is only a sample data and I can't see the original orientation in Axis Neuron Pro, but they are at the correct segment.
I added a new folder ("17N in Tristan's Office") with all the mentioned data to my drive. https://drive.google.com/drive/folders/ ... sp=sharing, after running the calc2csv.m file you just need to run the apdm_test.m file. my current transformation and heading correction you can find in the .xml files.
Could the problem occure, because the subject in the calibration data doesn't exactly match the default pose of the model?
I would appreciate any further help to this topic!
regards
Johannes