wrong results of BodyKinematics analysis

[Michał Wołek]

What may be the reason for the fact that BodyKinematics in AnalyzeTool calculates correct positions and velocities of CoMs and body orientations, but when it comes to accelerations, it does something strange and well... calculates it compeletely wrong. Pics attached

[Thomas Uchida]

when it comes to accelerations, it does something strange and well... calculates it compeletely wrong

It is unlikely that this type of calculation would be incorrect. My guess is that your position-level data are noisy. Differentiation degrades signal-to-noise ratio. Have you tried low-pass filtering before differentiating?

[Michał Wołek]

yes, the data is filtered, tried wwith different filtering frequencies, and changing angles to radians, none of which work. But noise isn't the problem here. Notice that despite velocity growing and falling down, the accelerations are constantly on the negative, and their value is huuge. Shape of the acceleration plot seems to be valid, but it looks like it got overscaled and translated by some huge amount to the negative.

[Thomas Uchida]

How were these results generated? Did you run the BodyKinematics Analysis with another OpenSim Tool?

[Michał Wołek]

I have used Analyze tool.

Anyway, I have found the solution to the problem, as other users have been also struggling with this and I found some old posts. BodyKinematics uses F=ma to get accelerations, instead of double-derivating positions. And since I'm trying to extract ground reaction forces from gait data alone, I don't any forces yet, that's why it gives wrong solutions.