Visual perception system for a social robot

This paper describes a visual perception system which allows a social robot to conduct several tasks. The central part of this system is an artificial attention mechanism which is able to discriminate the most relevant information from all the visual information perceived by the robot. This attention mechanism is composed by three modules or stages. At the preattentive stage, a set of uniforms blobs or 'pre-attentive objects' is obtained. Once the most salient objects are obtained, the semiattentive stage identifies and tracks some of them according to the tasks to accomplish. This tracking process allows to implement the `inhibition of return', avoiding revisiting an attended object. Finally, the attentive stage also fixes the field of attention to the most relevant object depending on the behaviours to accomplish. Three behaviours have been implemented which allow the robot to detect visual landmarks in an initially unknown environment and to recognize and capture the upper-body motion of people interested in interact with it.