一种基于2DLPP和2DLDA的人脸识别方法研究

局部保持投影（locality preserving projection，LPP）和线性鉴别分析（linear discrimin antanalysis，LDA）是两种有效的一维特征提取方法，广泛应用于人脸识别领域。但采用一维特征提取方法时会存在列向量化时样本的结构信息被破坏和样本在提取特征时必须对协方差矩阵进行特征分解，对于高维小样本的问题很容易出现协方差矩阵奇异的问题。文中提出将二维局部保持投影（2DLPP）和二维线性鉴别分析（2DLDA）这两种方法在特征层进行融合并应用在人脸识别。基于人脸库AR上的实验表明，该方法比传统的IJPP和LDA识别性能更高，因此可作为一种新的人脸识别方法。

Face Recognition Based on Feature Fusion by 2DLPP and 2DLDA

Locality preserving projections （LPP） and linear discriminant analysis （LDA） are two effective ID feature extraction methods, which have been widely applied to face recognition. However, such 1 D feature extraction methods always destroy the structure information in a face image when converting it into a vector. And since face images are high-dimensional, 1D methods suffer the singular problem （ small sample size problem ） when performing eigen-decomposition and inverse computation for the scatter matrices. In this paper,propose a novel feature fusion approach for face recognition, which fuses the features extracted by two-dimensional LPP （ 2DLPP） and two-dimensional LDA （ 2DLDA ）. The experiment based on AR face database shows that this proposed method can perform better results than the traditional LPP and LDA methods.