Measuring facial expressions by computer image analysis

Facial expressions provide an important behavioral measure for the study of emotion, cognitive processes, and social interaction. The Facial Action Co ding System (Ekman & Friesen, 1978) is an objective method for quantifying facial movement in terms of component actions. We applied computer image an alysis to the problem of automatically detecting facial actions in sequence s of images. Three approaches were compared: holistic spatial analysis, exp licit measurement of features such as wrinkles, and estimation of motion fl ow fields. The three methods were combined in a hybrid system that classifi ed six upper facial actions with 91% accuracy. The hybrid system outperform ed human nonexperts on this task and performed as well as highly trained ex perts. An automated system would make facial expression measurement more wi dely accessible as a research tool in behavioral science and investigations of the neural substrates of emotion.