Orientation offset?

[HWiese1980]

Hello @alvaroferran,

I'm currently trying to replicate your setup but I would like to have it as flexible as possible with regard to positioning sensors along the limbs. I'm wondering if and how you account and compensate for orientational offsets that occur when you place the sensor in a way that doesn't match the corresponding limb's reference frame.

Thanks!

[alvaroferran]

Hi, sorry for the late answer.
To be honest I didn't take it into account, I simply wore the sensors as shown in the model and then corrected small offsets by manually re-orienting the sensors using the model as feedback.

[HWiese1980]

I see. I'm trying to calculate and compensate those offsets automatically... not as trivial a task as one might think...

[ashbalaji-rehab]

Hi HWiese and alvaroferran,
I am facing the same issue trying to solve the orientation offset. Do you have any suggestions to resolve this issue?

[HWiese1980]

If I remember correctly (it's been a while since) I accounted only for the orientation offset around the yaw axis as the other orientations (pitch and roll) are absolute and can be used as is. I used another sensor source of information for calculating that yaw offset, namely an absolute skeleton tracking system that gave me an absolute vector from my elbow to my wrist at a rather low frequency. With that vector I was able to calculate a yaw angle in my desired coordinate system and compensate the yaw offset of the IMU with it. Worked like a charm, though not without the additional information. If you have other means to figure out the yaw offset, it should be all you need. Pitch and roll are absolute because they are calculated from the gravitational vector (provided you have an accelerometer in your sensor setup; I recomment BNO-055 IMU sensors, or their successors if available, for this purpose because they already do sensor fusion internally and are quite good and cheap).