Generating and Describing Affective Human-Agent Interaction

Lifelike agents, as a promising technology for human-computer interaction, have become focus of research community in resent years. In this paper, we will endow the lifelike agents with affective recognition capacity. There are three main contributions in this paper. Firstly, a hybrid of hidden Markov models (HMMs) and artificial neural network (ANN) is proposed to classify speech emotions. Secondly, a novel scheme for robust facial expression recognition is brought forward. Thirdly, the multimodal interaction markup language is designed to script web-based affective human-agent interaction, and we illustrate a scenario that instantiates the human-agent interaction described by MIML.