kinematics differences btw anybody analysis result and vicon nexus

Hi,

While I was looking into Anybody’s kinematics analysis result of upper extremities based on the motion capture data from the Vicon system, I found there are several mismatches of joint angle between the Anybody kinematics result and one of Vicon Nexus.

Especially, the data from the shoulder complex joint is severely varied.
First, I could get the shoulder joint angle by looking into
“Main.Studies.InverseDynamicStudy.Output.HumanModel.Right.ShoulderArm.Jnt.GH*.Pos”
but these kinematic results have very different patterns from the actual moving. I don’t think this parameter reveals the joint angle of GH joints; it might have another kinematic meaning.
Thus, I looked into another parameters
“Main.Studies.InverseDynamicStudy.Output.HumanModel.Right.ShoulderArm.InterfaceFolder.Glenohumeral*.Pos”
and I determined these results are the shoulder joint angle data.

However, when I compared these results from the kinematics data from Vicon Nexus, the joint angle is somewhat different.
If you see the attached excel files, the result at shoulder joint is very different each other. But the point is that the joint angle obtained from Anybody is not realistic.

Each tap has different arm motion, but the basic motion is the right arm is resting on the table first, moving to some point and being back to the resting position. The movement is like touching right ear, left ear, and the top of head.
We tried the simulation several time with some variation of marker number. Sometimes the head markers are removed, and sometimes the pelvic markers are removed with the same c3d file – just loading the markers or not by editing ModelSetup.any file.
The each motion of GH with different marker settings shows similar generally, but if I read the angle values, it shows about 1~2 radians which means impossible motion angle with normal person.
On the other hand, the joint angles from Vicon Nexus shows a moderate angle values which are agreeable.

I believe that there are surely result differences because there are difference in how to model the body and the joint and to define the joint coordinate in Anybody human model and Plug-in-Gait model in Vicon Nexus.
However, the results from Anybody is out of the acceptable range; that is where I cannot even imagine intuitively.

Can you see these attached excel files and how I can access those result reasonably?

Thank you.

Dear Sheen Dong-Pyoung,

You are right, the angles for the GH joint are found in the interface folder.

The angles in there don’t have to be the same as in Vicon. It is highly dependent what orientation the Vicon node of rotation has. So if one axis is different between Vicon and AnyBody, you will get huge differences in the resultant angles.

I had a look at the excel sheet and indeed the numbers seem very high in there, however, this does not necessarily mean that everything is wrong.
I would suggest to check a couple steps to see what happened:

  1. Does the motion look like it is supposed to look?
  2. In case you performed the motion and parameter optimization, how does the upper extremities look like? Are humerus, Ulna, and Radius in the right position initially?
  3. You can draw a ref frame inthe joint. make a fixed one in the joint origin and one that is moving with the humerus, so you can follow the motion better.
  4. Compare the motion of your C3D with eg. the FreePosture Model. you can easily insert into the Mannequin file the GH, Elbow and Wrist angles and see what happens there. Doe the postures look alike?
  5. If you know the orientation of the Vicon Reference Frame, you can setup a RefNode in the joint with this orientation and read out the angles in this RefNode…