基于Mutual Boost和支持向量机的表情识别

Facial Expression Recognition Based on Mutual Boost and SVM

主要研究自动人脸表情识别(FER),首先使用Gabor算法提取人脸图像的特征,再针对Gabor特征维数高、冗余大及利用传统的AdaBoost算法进行特征选择时特征间仍存在较大冗余的特点,引入了基于互信息的AdaBoost算法(MutualBoost)进行特征选择,降低特征维数和减少特征间的冗余信息量。然后再以SVM分类器进行分类。本算法在JAFFE表情库上进行测试,结果验证了算法的有效性。

Automatic facial expression recognition was studied.Facial image feature was picked up by using Gabor algorithm.For the high-dimensional Gabor feature vectors are quite redundant,Mutual Boost was introduced as a method of feature selection and redundancy exclusion.Then the SVM was used for classification.Experiments with JAFFE showed that the method is valid.