Mood estimation for human-robot interaction based on facial and bodily expression using a Hidden Markov Model

摘要

Understanding the emotion of an interlocutor is a critical human social skill. Thus, emotional assessment has become a subject of interest in studies of human-robot interaction (HRI). In this paper we propose estimating human moods by means of a Hidden Markov Model (HMM). This model assumes 1) there are only two hidden states (positive or negative mood), and 2) that these states can be recognized by certain facial and bodily expressions. The most significant parameter, face emotion score, is used to adjust the transition probability between these hidden states. A Viterbi algorithm has been adopted to predict the hidden state from the observed state of bodily expression. The model has been demonstrated in real time, showing that it could be used to enhance the skill of a social robot, thus endowing the robot with the flexibility to interact in a more human-oriented way.