Fisher's Discriminant and Relevant Component Analysis for static facial expression classification

This paper addresses the issue of automatic classification of the six universal emotional categories (joy, surprise, fear, anger, disgust, sadness) in the case of static images. Appearance parameters are extracted by an active appearance model(AAM) representing the input for the classification step. We show how Relevant Component Analysis (RCA) in combination with Fisher's Linear Discriminant (FLD) provides a good "plug-&-play" classifier in the context of facial expression recognition framework. We test this method against several other classification techniques, including LDA, GDA and SVM, on the Cohn-Kanade database.