Synchrony Detection as a Reinforcement Signal for Learning: Application to Human Robot Interaction

Abstract

The present study is aiming to build a synchrony-based attentional mechanism allowing to initiate and to maintain human robot interactions. Moreover, we question the importance of synchrony detection for learning and gaining new competences through the interaction. We previously proposed a synchrony-based neural model capable of giving the robot minimal abilities to select a human partner and to focus its visual attention on this preferred interactant. Here, we extend this model by using synchrony detection as a reinforcement signal for learning (during the interaction) the human partner appearance (shape) in the context of an autonomous mobile robot.