A hierarchical visuo-motor organization for computational models of the mirror system

Typical patterns in the space of motor configurations, frequently referred as motor synergies, enable one to provide a simplified action representation. The computational model of the mirror system introduced in this thesis hypothesizes that a synergy representation of actions could be actually developed in motor cortex and strongly involved in the process of elaboration of observed actions. In particular the action representation and the mirror system model were developed in order to reproduce two important biological findings about motor cortex, namely: 1) the action is represented according to different levels of detail; 2) the visual representation of an observed action is represented according to different levels of detail. In particular in this thesis, focusing on hand grasping actions, we investigate a synergy repersentation of actions where the synergies would be forced to be hierarchically organized. We show that this representation results to be accurate and, biologically plausible, being able to describe action at multiple level of detail. Successively we present an architecture where the previous hierarchical synergy representation is used in the process of elaboration of an observed action. We show how using this motor information can actually improve the process of action recognition and repreoduce one of the principal characteriscms of the mirror neurons, namely thier distinction in stricktly and broadly neurons.