I think a good approach would be automatic clustering of verts.
I.e for say animation 1, each vert generates a 'roadmap' of it's movement. Then group off verts that cross a threashold of similarity. (Say 90%)
Take this group, then average the roadmap of the verts applied to it, and then apply this to the overall limb.
Then any redudent verts are added to the group they most match, even if below the threashold..just to clean things up.
Won't be 100% perfect, but it should in theory(Unless I've had a brain shut down and can't see a glaring problem) work...
I.e for say animation 1, each vert generates a 'roadmap' of it's movement. Then group off verts that cross a threashold of similarity. (Say 90%)
Take this group, then average the roadmap of the verts applied to it, and then apply this to the overall limb.
Then any redudent verts are added to the group they most match, even if below the threashold..just to clean things up.
Won't be 100% perfect, but it should in theory(Unless I've had a brain shut down and can't see a glaring problem) work...