Different muscle activation depending on analysis start

Hi AnyBody Team,

I am uploading my AnyBody model here, in which I encountered difficulties concerning differences in the amount of muscle activity depending on when I start the analysis (Thread: http://forum.anyscript.org/showthread.php?t=3777)

There are six c3d files in the Input folder. The first number, is the subject ID and the second number the condition, number 3 and 4 do not matter.

It would be great, if you can explain me where those differences come from.


I attached 2 models, both for subject 13, with condition 1 and 2.

Here some changes, I changed marker positions, the foot area was definitely wrong before. You cannot analyse the first step, as there is no GRF available. I also added the shoulder markers, so the upper body gets some info as well. I still keep the manual drivers for the upper body as soft constraint.
I think VL is still active from around 5-10€ and peak is around 40% of stance.

Hi Amir,

thanks a lot for your help.
I am currently working through the model, to understand whcih changes were made and why.

How do I know, how to change the foot markers?

Also, I didn’t want to analyse the first step (since I have no force plate there), but I was wondering why the muscle activity of the step on the force plate changes depending on when I start my analysis.

If I am only intereset in stance phase. How many frames before and after HS and TO would you recommend to run the analysis?


The subject walked very strange in the trial you gave me, he had his toes upwards and a lot higher than the heel. Marker position can be edited in the marker.any file (sRel): the red marker on the skeleton needs to be where the real marker was on the subject. If you have a shoe you need to improvise a bit.

Thanks Amir,

it was a running movement, which might explain the foot position. And yes participants were wearing shoes.

I am comparing two different shoes (the two conditions), and get quite big differences in muscle activation of VL (with shoe 2 twice the activity of shoe 1). Those differences seem way to big for me (as in kinematics and GRF there are hardly any differences), therefore I was wondering what input parameter would cause those great differences?
Could it be that 5 frames before HS and after TO are not enough for a correct computation? If so, how many frames before HS and after TO would you recommend to run the analysis?


I ran subject 13_1_3_8 and switched between an start offset of 130 and 135. This leads to perfectly the same activation in VL.

If you still experience problems, please upload the model or explain stepwise how you generate those results.
Please also note, the step without recorded GRF cannot be used in this simulation. This might be the cause for the problem due to the inertia…

Do I understand this correctly, having the step without GRF in the simulation also influences the results of the step with GRF? leading to wrong simulations?


No, the step with GRF should not be influenced. That’s what I did and showed in the post above.

I created a new figure explaining what I mean (see attached):

I have two conditions (red and blue colour tone).

  1. I ran the simulation starting 5 frames before HS to 5 frames after HS (lighter colour)
  2. I ran the same files, starting the simulation with no start offset (darker colour)

For condition 1 (blue colour) results are very similar, but for condition 2 (red colour) results a quite different.

Therefore I had following questions:

  1. Why is there such a huge difference in muscle activation?
  2. Why do I see those great differences only for one condition, and not in both?
  3. Which input parameter might be responsible for this?
  4. Is there a recommended amount of time I should start the simulation (during running or walking) prior to ground contact?

I hope I was able to better present my question this time.

Unfortunately, I cannot reproduce your results. I get almost identical peaks for 13_2_3_4, no matter if I start at HS, 5 or 10 frames before.

Can you upload the two model-versions of subject 13_2_3_4 that bring different results.

Attached you can find the two models.

Only difference is in the Times.any file.
For the “full” model, I use the whole c3d file for the simulation (so starting 155 frames before HS).

I know that using the whole file including the step without GRF wouldn’t make sense, since it is not time efficient and since I can’t use the without-GRF-step. However, I found thos major difference in VL muscle activity calculation in the GRF-step and I am wondering where it comes from.

Thanks for looking in to this, Amir!

the motion in both models is different. foot on the right side is definitely wrong. The markers should not be under the foot, or?

you are right!

marker positions are the same in both files. so what did I do wrong for this to happen?

did you run the MotionAndParameterOPti again for this model?

yes, I did run MotionAndParameterOptimization for each trial.

No, seems like you did not run the Opt study after my changed marker settings from last week. I just rerun it and it looks ok now.

I had a look again. I ran them for sure. The OpitimizedParameters.txt and the other txt files were from Dec 18th the day when I uploaded the model. And I just checked, and it looks fine in my model.
Sorry, I didn’t check my model again, after you posted the figures, since I assumed that I had an mistake.

However, my main question still remains, why is the amount of muscle activation during the GRF-step different in those two models?

Thanks for a lot your help!

It seems, that the kinematics are different if you analyze a longer trial. Diffferent kinematics will lead to different forces/activations. I will look at the model again.

Hi Amir,

I hope you had a great start into the new year!
I wanted to ask, if you already had a chance to look at the model again?
Any idea, why the kinematics are different in a longer trial? and how to best set up my analysis to get the correctest kinematics?


There is a different kinematic solution depending how long the trial is. This will also affect muscle activation in the inverse study, so you need to make sure the kinematic solution is ok.