Occurrence order detection of face features deformations in posed expressions

Abstract

Timing, duration, speed and occurrence order of facial actions are different parameters related to dynamic analysis. These parameters are crucial in human behavior understanding and in distinguishing posed from spontaneous expressions. In this paper, the occurrence order parameter is analyzed. Motion information is extracted from image sequence by tracking characteristic facial points. Extracted information is got in the form of multi time series. Allen's Logic Algebra is used to model the occurrence order parameter in these time series. A rule based method is used to learn order rules. These rules are evaluated in the classification of new examples into one of five facial expressions namely: Joy, Disgust, Anger, Sadness or Unknown. The system is evaluated in terms of recognition accuracy. High rates have been obtained when classifying new unseen examples. © 2011 Springer-Verlag.