This add-on follows closely the idea of the existing BVH importer for motion capture data: Given an MVNX file, it loads the joints of the described skeleton into a Blender Armature, and then loads the positions and orientations of the given sequence.
At the moment only import functionality is supported, since it is lossy (MVNX is much richer than BVH) and exporting back would require lots of extra informations. ATM it is unclear to me how to generate them from scratch or how to import/infer them all without losing code flexibility/utility/maintainability. As it is, the addon is clear and extendable. it will help my team to bypass the proprietary rendering software by XSENS and presumably/hopefully could help others out there. I will be happy to hear about bugs and fix them.
The addon itself is in the "io_anim_mvnx" folder:
I tested it on different MVNX versions and compared the imported results with corresponding BVH ones.
I also tested all the combinations for the property fields.
The repository hosts an autodoc as well as unit test facility with some data examples (I didn't cover much with the utests though).
It passes flake8 and tried my best to document it well.
So far the addon is small and I was lazy, so I didn't set up the Travis CI to have it running on Blender (if you have some copypaste solution let me know and I will incorporate it).
Please let me know if there is something to improve, I will be glad to contribute to Blender. Thanks for your time and effort!