Classification of spontaneous and posed smiles

Automatic detection of affective states from facial images and videos is a very important application in social signal processing. In this study we focus on the expression of happiness, and propose a physiologically inspired approach for detecting whether a smile is spontaneously produced or not. We use distance-based and angular features that characterize movements of facial regions, and train local classifiers for smile classification. Our experiments with different classifiers on two different datasets show that the eyelid region can be very useful in assessing spontaneous versus posed smiles.