Simultaneous Recognition of Face and Facial Expression Via Partial Least Squares Regression

doi:10.11834/jig.20090507

Views : 0 下载量: 293 CSCD: 0

PDF
Export
Share
Collection
Album

Simultaneous Recognition of Face and Facial Expression Via Partial Least Squares Regression
Vol. 14, Issue 5, Pages: 801(2009)
Published：2009
DOI： 10.11834/jig.20090507
稿件说明：

移动端阅览

DOI：

Simultaneous Recognition of Face and Facial Expression Via Partial Least Squares Regression[J]. Journal of Image and Graphics, 2009, 14(5): 801. DOI： 10.11834/jig.20090507.

摘要

将偏最小二乘回归方法用于人脸身份和表情的同步识别。首先，对每幅人脸图像进行脸部特征提取以及相应的语义特征定义。在脸部特征提取方面，从每幅图像中标定出若干脸部关键点位置，并提取图像在该关键点处的Gabor小波系数（Gabor特征）以及关键点的坐标值（几何特征），作为该图像的输入特征。语义特征则定义为该人脸图像所属的表情类别信息以及所对应的人脸身份信息。其次，利用核主成分分析(KPCA)方法对脸部Gabor特征和几何特征进行融合，使得输入特征具有更好的识别特性；最后，运用偏最小二乘回归(PLSR)方法建立脸部特征和语义特征之间的关系模型，并运用此模型对某一测试人脸图像进行表情和身份的同步识别。通过在JAFFE国际表情数据库和AR人脸数据库上的对比实验，证实了所提方法的有效性。

Abstract

In this paper

we utilize the partial least squares regression (PLSR)method to solve the simultaneous recognition problem on face identity and face expression. Firstly

face features and the corresponding semantic features are extracted for each face image as the input features

where the face features are defined as the Gabor wavelet coefficients defined on several land marks of each face image;the geometric features are defined as the coordinates of the landmark points and the semantic features are defined as the facial expression category index and face identity index of each face image. The kernel principal component analysis (KPCA)method is then applied to deal with the feature fusion task of both Gabor features and geometric features. Finally

the PLSR method is used to model the correlation between the input facial feature vectors and the semantic vectors. Based on this model

both face identity and facial expression category of any test facial image can be predicted. Experiments on both JAFFE facial expression database and AR face database show the effectiveness of the proposed method.