Exploring the space of emotional faces of subjects without acting experience

The current state of semi-automated facial animation demands better understanding of how facial expressions are produced by real people. As a first step of a series of empirical studies, we investigated snapshots of facial expressions of the six basic emotions, produced by 18 subjects in `out of context' sessions. The MPEG-4 coded vectors were analysed on principal components and canonical variates, which served as a basis to draw conclusions about the (for the negative emotions: lack of) generic characteristics of the six emotions and their mutual distances. Our results are compared to ones obtained by facial image analysis and human perception researchers.