人脸活动单元自动识别研究综述赵晖.pdf

22 5 Vol.22 No. 5 2010 5 Journal of Computer-Aided Design Computer Graphics ay 2010 1,2) 2) 2) 赵 晖 王志良 刘遥峰 , , 1) ( 830046) 2) ( 100083) (zhaohui@xju. edu.cn) : ( action units, AU), . AU , AU ,AU AU AU AU .AU ,. : ; ;; : T P391. 41 A Survey of Automatic Facial Action Units Recognition 1,2) 2) 2) Zhao Hui , Wang Zhiliang , and Liu Yaofeng 1) (School of Inf ormation Science Engineering, X inj iang University, Urumqi 830046) 2) (School of Inf ormation Engineering, China University of Science Technology, Beij ing 100083) Abstract: Automatic facial action units ( AU ) recognition can be applied in many areas such as behavioral science, man-machine communication, security and medical diagnosis, and has drawn much attention recently. This paper explains the conception and general procedure of automatic AU recognition, and introduces the main methods of feature extraction and AU classification. Besides, several typical AU-coded facial expression databases are presented, and the state of single AU, AU combination, AU temporal dynamics andAU intensity recognition is demonstrated. Finally, this paper summarizes the current difficulties of automatic AU recognition, and outlines the future development trend. Key words: facial action units; facial action coding system; feature extraction; expression recognition ( action unit, AU)Ekman , [2] ,20 . 70 (facial action ,, [ 1] coding syste