Automatic Facial Expression Recognition Using both Local and Global Information

A novel approach to automatic facial expression recognition from static images is proposed in this paper. First, active appearance model (AAM) is used to locate facial feature points automatically. Then, both local texture information and local shape information in these points are extracted and are combined with global texture information for face presentation. Finally, the linear programming (LP) technique is used for classification. Experimental results demonstrate an average recognition accuracy of 83.6% on the JAFFE database, which shows that the proposed method is promising.