Just to clarify: I am using action to denote chunks of movement:
Action is thus highly subjective, it is just a mental construct (for either the performer or perceiver, or both) of chunks in the continuous flux of movement. Acknowledging the fact that our brain is working at multiple speeds and resolutions, there could also be actions that are chunks of smaller actions.
It is these actions (i.e. movement chunks) that will be the basis for gestures (i.e. actions with semantics).