Differentiating Duchenne from non-Duchenne smiles using active appearance models

Face-related biometrics research in recent years has moved from attempting merely to recognize faces, and even doing so under varying conditions, to considering a wide variety of aspects such as dynamics, gesture, aging, and expression. The state of an individual's face is a revealing indicator that may be used for soft biometrics, active authentication, deception detection, response feedback, and other areas of interface. One related psychological indicator is the Duchenne smile that usually indicates a genuine, spontaneous, or enjoyed emotional state rather than a forced or posed state, as likely expressed by a non-Duchenne smile. Differentiating between these is a useful task to automate for a variety of reasons. This paper discusses a classification technique that achieves higher recognition rates than previously published for similar comparisons.