Filtering Data - Follow-Up

[Herman van Werkhoven]

Hi Nick,

Thank you for your response on my initial post on Nov 20. I just wanted to get more details related to the filtering. We are comparing results from OpenCap to AddBiomechanics generated joint kinematics and the filtering details are important. I would really appreciate it if you can answer a few more questions:

• Is it suggested to filter input marker data? I've always thought that filtering should occur as early in the process as possible, but maybe not?

• You wrote regarding AddBiomechanics filtering - 'Filtering is applied after the kinematics step (e.g., marker fitting) to smooth joint kinematics'. Does this mean that the joint angle kinematics outputs (in the *.mot) files are filtered, or does the filtering only occur after this step, and really only affects dynamics results?

• If the joint angle outputs are filtered, where can I find details of the filter method (e.g., type of filter, cut-off frequency etc.)?

Thank you,
Herman

[Nicholas Bianco]

Hi Herman,

Is it suggested to filter input marker data? I've always thought that filtering should occur as early in the process as possible, but maybe not?

That is mostly up to you. If you believe that your marker data has noise that may affect the kinematics fitting step, then you may apply filtering. It is most important that each marker trajectory in your data 1) matches a marker in the uploaded model, 2) is "clean" (e.g., free of artifacts, significant noise, gaps, etc.), and 3) does not swap with another marker (which is a very common problem). We try to address as many of these issues in our heuristics, but the more clean your input data is, the better your results will be.

Does this mean that the joint angle kinematics outputs (in the *.mot) files are filtered, or does the filtering only occur after this step, and really only affects dynamics results?

I'm not sure what you mean exactly, since the output of the kinematic fitting step is the trajectory of joint angles, which get filtered before going to the dynamics fitting step, i.e., kinematics fitting --> joint angle filtering --> dynamics fitting.

If the joint angle outputs are filtered, where can I find details of the filter method (e.g., type of filter, cut-off frequency etc.)?

We've been using the term "filtering", but the latest version of AddBiomechanics does something a bit different: we perform a least-squares optimization to minimize the acceleration of each joint for the while also tracking the original joint angles closely. The details of this process can be found in the =https://link.springer.com/chapter/10.1 ... ublication.

Best,
Nick