Compute joint angles

[niksirbi]

We have always wanted to implement functionalities for joint angles in movement, but I realised we don't have an issue to track that.

I'm opening this broad issue to start the discussion on joint angles:

• how should we represent angles (Euler, quaternions, etc.)
• What should be the functional interface for computing them?
• Where should these functions live (probably in movement.kinematics)?

Useful resources

• Discussion about joint angles in freemocap, thanks for the shoutout @roaldarbol
• Motion-BIDS paper: includes a spec for specifying angles in Euler or quaternions
• pyomeca—an xarray-based biomechanics package: includes Angles data arrays.
• compute_joint_angles() function in DLC2Kinematics

Feel free to comment with any other resources that come to mind.

Related issues

Some of these are only tangentially related, but I'm listing them here for completeness/findability.

• Add I/O support for Motion-BIDS #581
• Add the ability to import skeletons #93
• Computing a (smooth) angular head velocity over time #367

Note

This feature is at the discussion stage, not yet ready to be implemented. Feel free to contribute to the discussion!

[vtushar06]

Hi @niksirbi!

I'm interested in contributing to this joint angles discussion. This looks like a really valuable feature for the movement package.

Since this is at the discussion stage, could you guide me on how I should start? Should I begin by:

• Reviewing the resources you've linked (pyomeca, Motion-BIDS paper, etc.)?
• Looking at the existing movement.kinematics module structure?
• Researching specific use cases for joint angle computation?

I'd appreciate any direction on where to focus my initial efforts!

Thanks!

[niksirbi]

I think the best starting point would be to review the resources I've linked and see if there is any common representation for joint angles that could be abstracted away. Keep in mind that our representation has to be general enough to work for any species, we cannot afford to make any human-specific / mouse-specific etc. assumptions.

[roaldarbol]

Just some resources.

https://www.researchgate.net/post/Does_anybody_know_if_there_is_some_comparative_technical_report_about_the_use_of_Euler_angles_vs_Quaternions_in_human_motion_analysis

https://pmc.ncbi.nlm.nih.gov/articles/PMC7422562/

I think my preference would mirror that of the ResearchGate discussion. I think quaternions are less interpretable than Euler angles - simply because people don’t know them, and you’d need to know a thing or two about complex numbers. But if rounding errors are an issue, then internally converting to quaternions and return in Euler. Would require a bit more work, but could avoid the issues related to Euler angles.