基于双正交小波和加权NLDA的人脸识别

文章针对基于零空间线性判别分析方法在高维多类别的人脸识别应用中存在的“次优性”和“效率低”两个问题,提出了一种改进方法.该方法用加权函数重新定义类间离散矩阵以减弱边缘类带来的不良影响来改善次优性问题；并对图像进行双正交小波分解来降低分辨率以到达提高效率的目的.最后在ORL人脸库上进行了实验,通过实验确定了最优加权系数和小波分解的最佳层数,并验证了该方法比基于零空间线性判别分析方法有更高的效率和识别率.

Face Recognition Based on Biorthogonal Wavelet and Weighted NLDA

In order to improve "sub-optimal" and "low efficiency" problems of based on null space LDA(linear discriminant analysis) in high-dimensional and multi-class face recognition application,a method is proposed in this article.This method used a weighting function to redefine between-class scatter,that weaken the edge-class’s adverse effects,and this method also used biorthogonal wavelet to decompose the images to reduce the resolution,that could improve the efficiency.At last,did experiments in the ORL face database,which determined the best weighting factor and the best layers of wavelet decomposition,at the same time,the experiment verified this method dose have a higher efficiency and recognition rate than based on NLDA.