High frequency noise in SO results after RRA.

[Daniel Pina]

Hello,

I have some lab results consisting of gait tests with camera markers and pressure plates. I did all the scaling and inverse kinematics procedures and they seem to be correctly done.
Then I ran RRA for each test, after which I ran SO to study muscle activation.
I used the gait2392 default files for tracking tasks and reserve actuators, and for SO I used the gait2392 default file for CMC actuators. The only thing I changed was the coordinate of action for the point actuator FX/FY/FZ, to the mass center of my subject's pelvis.

Here are 3 samples of the SO results I'm getting for muscle activation on a gluteus muscle in each leg and hip extension/flexion residuals after the procedure:

Gait_Test01.png (29.45 KiB) Viewed 683 times

Gait_Test02.png (28.04 KiB) Viewed 683 times

Gait_Test03.png (27.99 KiB) Viewed 683 times

Is this noise/jitter typical of SO results? The samples I got from the tutorial didn't behave this way.
I tried to increase the optimal force of the reserves in the gait2392_CMC_actuators file from 1 to 5 and then to 50, but I didn't get much different results.

One thing I noticed is that running RRA returns motion files with a 1000Hz sample rate, whereas my tests were made at 200Hz. This means that for SO, I used the RRA "kinematics_q" file as motion file at 1000Hz together with a GRF at 200Hz. Could this be part of the issue? I noticed the console window mentions it's using a GRF file with the "transformedP" suffix. Does this mean OpenSIM is automatically upscaling the frequency of my GRF file, or do I have to do that manually?


Thank you so much for your support.

[Daniel Pina]

A little update.
After studying the documentation again, I noticed I hadn't locked the MTP and subtalar angles for RRA.
After this procedure in Test01, I got somewhat cleaner results in SO:

Gait_Test01_locked_mtp_subtalar.png (24.09 KiB) Viewed 644 times

However, there's still a lot of noise.
Since all the motion results from RRA present very smooth curves, I went to check the GRF curves because those are introduced in SO as well. Some components in my GRF files are very noisy:

Gait_Test01_GRF_Torque.png (21.29 KiB) Viewed 644 times

Gait_Test01_GRF_Force.png (22.03 KiB) Viewed 644 times

(I left out the "Y" components in these graphics because they're smooth and their value ranges are superior which would take definition from the other signals in the picture)


Could my GRF files be the culprit for the noisy signals and the muscles are trying to compensate for this noise?
Should I run these values through a low-pass filter, for example? This is rather strange because I'm checking the "low-pass filter" box for 6Hz in the GRF xml file.
Or could all this be from using motion files resulting from RRA at 1000Hz together with a GRF file that stands at 200Hz?
If so, is there a way to reduce the frequency output of RRA to the same 200Hz as my GRF file?


Thank you.

[Daniel Pina]

A little update, but I'm still going around in circles so some help would be greatly appreciated.

I went to check the SO force file and the "transformedP" GRF values are there. It seems OpenSIM not only upscales the frequency from the 200Hz initial value to RRA's matching 1000Hz, but it also seems to filter those values as the transition is smooth and there's no aliasing.
So I don't think this high-frequency noise is the fault of a GRF-RRA mismatch in frequency.

[Daniel Pina]

Update:
I have solved my issue and it was indeed because I had noise in the GRF files.
Here's the thread where I explain it better:

viewtopicPhpbb.php?f=91&t=8289&p=22774&start=0&view=