Automatic facial expression interpretation: Where Human-Computer Interaction

Recently there has been an increasing incentive from researchers in Human-Computer Interaction (HCI, henceforth) – who study the mental and physical world of computer users, develop approaches and methods to predict properties of the interactions and to support the design of interfaces – to perform a fundamental shift in the way they think about input-output interactions with a computer: a shift toward a human-centered interaction architecture, away from a machine-centered architecture. The main motivating principle is that computers should be adapting to people rather than vice versa. Since faces are at the center of human-human communication, it would seem natural and desirable to find faces at the center of human-computer interaction. Indeed, there has been considerable technical progress within Artificial Intelligence (AI, henceforth) – an area of research which has evolved to develop “intelligent agents ” applying machine learning techniques to help people interact with computers – in the field of computer vision to open the possibility of placing faces, if not at the center of human-computer interaction, at least at a significant place within man-machine interaction. It still seems to be the case, however, that computer vision researchers, interested in how to solve problems, are working in isolation to develop ever more sophisticated algorithms to recognize and interpret facial information without necessarily knowing what they would do with the information, were they to solve th