一种新的模糊对称散布子空间的鉴别分析算法

New discriminant analysis algorithm using fuzzy symmetrical scatter subspace

提出一种模糊对称散布子空间的鉴别分析算法.首先通过构造对称散布子空间模型,分别获得类内和类间散布矩阵的一组子空间;其次引入了一种松弛归一化条件的模糊鉴别分析算法,根据每一个样本的隶属度对散布矩阵重定义所做的贡献,将它融入到特征抽取的过程中,从而得到完整有效的模糊特征向量集,解决了传统LDA方法中的最终特征维数受类别数限制的问题.在NUST603和ORL人脸数据库上的实验结果验证了该算法的有效性.

A novel fuzzy discriminant analysis（FDA） on the symmetrical scatter subspace（SSS） was proposed in this paper.First,an SSS model on the discriminant analysis was established,by which a set of integrated subspaces of within-class and between-class scatter matrices were constructed,respectively.Second,a reformative fuzzy LDA algorithm based on the relaxed normalized condition was introduced to achieve the distribution information of each sample represented with fuzzy membership degree,which was incorporated into the redefinition of the scatter matrices.The problem,in which the final dimension of features obtained by Fisher′s discriminant analysis was confined by the number of classes,was solved.Extensive experimental studies conducted on the NUST603 and ORL face images show the effectiveness of the proposed algorithm.