用隐马尔可夫模型设计人脸表情识别系统  

Design of Recognition for Facial Expression by Hidden Markov Model

作　　者：尹星云 王洵[1] 董兰芳[1] 万寿红[1] 

机构地区：[1]中国科学技术大学计算机科学技术系

出　　处：《电子科技大学学报》2003年第6期725-728,共4页Journal of University of Electronic Science and Technology of China

基　　金：安徽省自然科学基金资助项目;编号:2003KJ293

摘　　要：根据隐马尔可夫模型(HMM)的基本理论和算法设计了一个人脸表情识别系统。该系统由两层HMM组成:低层由六个HMM组成,分别对应六种特定表情。人脸表情特征向量进入系统后,经过低层HMM初步识别,其结果组成高层HMM的观察向量,经过高层HMM解码,确认出表情,从而提高了系统的识别率,增强了系统的健壮性。Hidden Markov Model(HMM) is a widely used statistical model.This paper deal with the design of a recognition system for facial expression in the light of the principal theory and algorithm of the HMM.The system consists of two levels of HMM,with the low level composed of six HMM,corresponding to six expression-specific.The vectors of the facial expression features,after put into the system,is primarily identified through the low HMMsand result in the observation vectors of the high level of the HMM.Through deciphering of the HMM,the expression are identified,thus enhancing the recognition rate and strengthening the system to a higher level.

关 键 词：隐马尔可夫模型 forward-backward算法 VITERBI算法 Baum-WeUch算法 人脸表情识别 人脸表情特征向量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]