A Story-Telling Social Robot with Emotion Recognition Capabilities for the Intellectually Challenged

Abstract

In this paper, a story-telling social robot is proposed. The robot is able to modify the evolution of the story considering the emotions the audience is feeling. To do that, the robot uses the user’s emotion from his/her face. We have used a deep learning-based model to identify the emotion. This model was trained and tested on state of the art dataset. We also demonstrate that involving generated, realistic samples in the training process as a way of data augmentation does not benefit the model at all. The mentioned samples are generated by a state of the art GAN, which is able to translate neutral faces to a range of emotions.