The goal of HCI researchers is to make interaction with computer interfaces simpler, efficient and more natural. In a context of object manipulation, we think that reaching this goal requires the ability to predict and recognize how humans grasp then manipulate objects. This is based on studies explaining human vision, reach, grasp taxonomies and manipulations. In this paper, we study the full cycle of gesture interaction using different points of view, then attempt to organize them using Norman's theory of Human Action, we link the psychology of object sensing to HCI goals and propose a simplification of gestures classes into four principal families. Our simplification of gestures classes still allow the expression of more detailed subclasses differentiated by the gesture properties.