In the field of 3-D action recognition, self-supervised learning has shown promising results but remains a challenging task. Previous approaches to motion modeling often relied on selecting features ...