Recognition of facial expressions using muscle-based feature models

Abstract

We present a technique for recognizing facial expressions from image sequences. The technique uses a muscle-based facial model for tracking motion of facial components, such as eyebrows, eyes, and mouth. This model consists of facial feature points and vectors corresponding to facial muscles. The contractile degrees of facial muscles are obtained from the deformations of the model by matching the model with facial images. Plausibilities of facial expressions are defined between those muscular contractile vectors and representative vectors of principal expressions. Experimental results show that those vectors correspond to the changes of facial expressions and that four facial expressions, happiness, surprise, anger, and sadness, can be recognized by the vectors. By thresholding the plausibility and finding the beginning and the end of a certain expression, an image sequence can be partitioned into segments of expressions. © 2000 Scripta Technica, Syst Comp Jpn, 31(10): 78–88, 2000