The Future of Therapy: A Study on Emotion Recognition Solutions Using NAO Robot

Behavior adaptation and personalization are critical in human-robot interaction. Thus, automatically collected feedback from people proves to be a feasible solution. Human emotions represent insightful feedback that can be collected from people. This paper presents a comparative study of some well-known emotion recognition neural networks currently available. To analyze network performances, a new dataset was created using images obtained from the humanoid robot NAO. Furthermore, we studied the impact of gender separation on the accuracy of the emotion recognition neural networks.