基于组合方法的人脸识别算法研究

该文试探一种加权融合流形学习的方法进行人脸识别，该算法通过Haar小波和局部线性嵌入（LLE）加权融合的方式来进行人脸识别。首先通过Haar小波变换对人脸图像进行一级分解，得到四个子图；然后利用LLE算法对四个子图降维处理，并加权融合；最后利用支持向量机（SVM）进行分类判决。通过在ORL库中进行实验，结果表明，该文提出的Haar＋LLE识别效率比单独使用主成分分析（PCA）和LLE更高效。

Face Recognition Based on Combination Method

A novel algorithm based on manifold learning is explored for face recogni-tion.The algorithm used method of Haar wavelet and Locally Linear Embedding（LLE） of weighted fusion for face recogni- tion.Firstly,the face graph is gone to be one level decomposition by Haar wavelet transform and become four subgraphs;Secondly,the four subgraphs of dimensionality are re-duced by method of LLE;Finally,the features are classifyed and recognized by SVM.Experimental results on the ORL face data-base show that the new algorithm has a recognition ability than the algorithm of PCA and LLE.