Neural approach for personalised emotional model in human-robot interaction

Robotic technology is widely spreading into our everyday lives so the social interaction between robots and humans is becoming more important. We believe that the quality of human-robot interaction will determine the effectiveness of the collaboration and in general, the acceptance of robots in the society. With this motivation we propose a more intuitive way of interacting with robotic companions by recognizing human emotions and, on the other side, a methodology to construct a comprehensible way of expressing internal states of machines for human users. The novelty of our model consists in personalization of the emotional expressions for each of the users. The system consists of two parts: learning emotional expressions of humans using ARTMAP neural network and implementation of the personalised emotional model to a humanoid robot. This paper shows the first results based on experiments realized on the humanoid platform Nao. We achieved a personalised emotional behaviour especially useful in the area of social robotics.