Mapping emotional status to facial expressions

The facial expression plays a crucial role in interpersonal communication. We can perceive people´s inner world by observing his/her facial expressions. We present a simple methodology for synthesizing realistic facial expressions by manipulating emotional status. For the convenience of parametric representation, we adopt a statistical model to describe facial appearance variations due emotional factors. We investigate the correlation of parameters between emotional status and face model, and design an emotional function that maps one to another. Since the emotional function considers the way in which pose expressions only, it is independent of the particular subject within the training set and can be used to simulate emotional expressions for a new person. The capability of emotional function enables us to conduct a series of interesting experiments, such as predicting unseen expressions for a unfamiliar person, simulating one´s facial expression in other´s style, extracting pure expressions from a mixture. All these experimental methods and visual results are presented.