基于加权邻域极大边界准则的人脸识别

摘要

The maximum margin criterion (MMC) is a linear feature extracting algorithm which computes an optimized projection by solving the generalized eigenvalue problem in a standard form that is free from inverse matrix operation. And thus it does not suffer from the small sample size problem. However, MMC preserves the global structure of dataset to obtain the projecting matrix. In face recognition, a local geometry structure is essential for classification. Hence, an improved MMC, namely weighted neighborhood maximum margin criterion (WNMMC) is proposed. Unlike MMC, WNMMC preserves the local geometric structure of database. The objective function of WNMMC leads to the enhancement of classification capacity by using the linear reconstruction coefficients. To verify the efficiency of the proposed algorithm, experiments are conducted on two face database. Experimental results show that the method consistently outperforms other recognition methods based on principal component analysis (PCA), linear discriminant analysis and maximum margin criterion.%极大边界准则是近年来提出的一种有监督的线性空间降维方法,该方法通过求解一般的特征方程来获得最优的特征向量,不用计算高维矩阵的逆,克服了特征提取中遇到的小样本问题.然而,极大边界准则只选择数据的全局结构,忽略了数据局部几何结构,而在人脸识别中,数据的局部几何结构起着非常重要的作用.针对极大边界准则这一局限性,提出了一种新的极大边界准则算法.该方法选择数据的邻域点最优重构系数用在目标函数中,保留了数据的局部几何结构,从而在低维空间中提取出更好的分类特征.本文还将该方法用在人脸识别中,通过在两个数据库中的实验,证明了其较主成分分析法,线性判别式方法以及平均邻域极大边界准则算法具有更好的识别性能.