Physiology-based affect recognition and adaptation in human-machine interaction

Abstract

Recent advances in robotics and intelligent systems are expected to usher in a new era where the need for machines to “understand” humans becomes increasingly important. It should permit more meaningful and natural human-machine interaction (HMI) when a robot/computer can detect the affective cues of the person it is working with. The objective of this work is to investigate the following hypotheses for achieving an affect-sensitive HMI: (i) It is possible to detect the affective states of interest by using multiple indices derived from physiological signals in real-time; (ii) Such affective cues can be integrated within a machine's control architecture to make it capable of responding to them appropriately; and (iii) Such affect-sensitive systems are expected to improve the overall human-machine interaction experience. In this work, a systematic comparison of the strengths and weaknesses of machine learning methods was performed when they were employed for the physiology-based affect recognition. The impacts of the affect-sensitive closed-loop interaction were investigated in both human-robot interaction (HRI) and human-computer interaction (HCI) contexts. Furthermore, in response to the growing need for developing robot/computer assisted autism intervention systems for children with autism spectrum disorder (ASD), physiology-based affective modeling and adaptation methods were investigated for this specific population. Finally, physiology-based affective modeling using active learning for children with ASD was discussed.