一种新的人面部表情识别方法

New Method of Facial Expression Recognition

论文提出一种全自动识别人脸七种基本表情(愤怒、厌恶、恐惧、高兴、无表情、悲伤和惊奇)的方法。该方法首先采用几何模型匹配法自动定位出人眼,在此基础上进行人脸大小归一化,然后利用局域二值模式(LocalBinaryPattern,LBP)技术提取面部特征,最后采用由粗到细的方案进行表情分类。采用日本的JAFFE公用表情数据库来检测算法的性能,实验结果验证了方法的有效性。

In this paper,we present a new method for automatic facial expression recognition.We analyze seven basic expressions ： angry,disgust, fear, happiness,neutral,sadness and surprise.First,locate the eyes automatically with the method of geometrical matching;then the Local Binary Pattern（LBP） operator is used to extract face appearance features;finally a two-stage classification method is proposed.The algorithm is tested on the JAFFE database and the result approves its feasibility.