Geração de apresentações orais com expressividade emocional em um robô humanoide

Abstract

With the increase in activities performed remotely, self-service systems and, more recently, the need for social distancing due to the COVID-19 pandemic, the use of robotics and autonomous systems for activities that demand interactions with the public have been increasingly studied. Among the emerging solutions to this demand are social robots and, in particular, humanoid robots. The latter, through their anthropomorphic form, seek to communicate with humans in a more intuitive and natural way, causing a sense of social presence, thus achieving a better engagement with people. Within this context, an extremely relevant problem is the expression of natural gestures and movements, which are in tune with the robot’s speech. The act of demonstrating emotions in speech increases affective communication and understanding of what is said by generating empathy between the robot and its target audience. In this work we propose a method for the automatic generation of oral presentations, from written scripts, with body movements and expressions, based on emotions automatically associated with the text, via algorithm trained with artificial neural networks.