中文摘要：

为了提取人脸图像丰富、有效的互补特征集,建立三种基于空域、频域和u域（分数阶傅立叶域）的特征提取模型,分别为基于局部二元模式（LBP）的空域多分辨率特征提取模型与基于频域和u域混合特征提取模型。在决策层,用加权和的方法对三种模型得到的相识度矩阵进行融合得到总的相识度矩阵,用最近邻分类器进行分类得到识别结果。实验表明,该方法能提取出丰富、有效的判别特征,与基于单一特征形式的人脸识别方法相比,识别效果得到了较高的改善。

英文摘要：

To extract rich and effective complementary feature sets of face images,three kinds of feature extraction models are built based on spatial domain,frequency domain and u domain（ fractional Fourier domain） respectively,they are the multi-resolution feature extraction model based on local binary patterns（ LBP） for spatial domain,the hybrid feature extraction models for frequency domain and u domain.On decision level,the weighted sum rule is used to fuse three similarity matrices generated with three models to derive general similarity matrix,and then the nearest neighbour classifier is used for classification to get the recognition result.Experiments show that this method can extract rich and effective discriminant features.Compared with face recognition method based on single feature extraction process,its recognition performance is improved noticeably.