Convolutional neural network oriented to the gesticulation control of an interactive social robot with humanoid aspect

Abstract

Social robots present the boom of an infinity of investigations in terms of design and control, these in turn approach with greater intensity to be our partners at home or work, however, still lack similar social interaction skills to the human being. This paper presents the implementation of a convolutional neuronal network algorithm aimed at the arrest of emotions for decision-making in the gesticulation of the social robot with a humanoid appearance. In addition, emphasis is placed on the use of visemas as a unit of labial movement for improve the human-robot interaction in any field in which the social robot can be used, be it teaching, therapies, safe navigation, etc., simultaneously coordinating and similar to the human naturalness the movement of servomechanisms of the mouth and the synthesized audio of the text that the robot gives in response.