Wearable vision based handled objects classification in human robot interaction

Human robot interaction highly desire natural and human-Oriented way in order to enable human and robots better work together, so robot vision should understand human or self activities in egocentric point of view. In this paper, we present a method for visual classification of objects in human hand captured from a wearable vision. The method tackles some challenges of object classification in egocentric perspective; Knowledge of the user's hand location and shape context assists the objects classifier based on previously learned in a labeled set of context. We present experiment on a dataset sellected within a cluttered environment, implementation validates its validity.