基于LDA算法的人脸识别方法的比较研究

A Comparative Study on Face Recognition Using LDA-Based Algorithm

线性判别分析(LDA)是一种较为普遍的用于特征提取的线性分类方法。但是将LDA直接用于人脸识别会遇到维数问题和“小样本”问题。人们经过研究,通过多种途径解决了这两个问题并实现了基于LDA的人脸识别。文章对几种基于LDA的人脸识别方法做了理论上的比较和实验数据的支持,这些方法包括Eigenfaces、Fish-erfaces、DLDA、VDLDA及VDFLDA。实验结果表明VDFLDA是其中最好的一种方法。

Low-dimensional feature representation with enhanced discriminatory power is of paramount importance to face recognition （FR） system. Linear Discriminant Analysis （LDA） is one of the most popular linear classification techniques of feature extraction, but it will meet two problems as computational challenging and “small sample size” when applying to face recognition directly. After studying people solve the two problems through several ways and realize the face recognition based on LDA. The short paper here makes compare on theory and experimental data analysis on several Face Recognition system using LDA-Based Algorithm, such as Eigenfaces （using PCA）, Fisherfaces, DLDA, VDLDA and VDFLDA. The experimental results show that the VDFLDA method is the best of all.