提出一种新的基于典型相关性的核鉴别分析,以图片集为基础的人脸识别算法。把每个图片集映射到一个高维特征空间,然后通过核线性鉴别分析(KLDA)处理,得到相应的核子空间。通过计算两典型向量的典型差来估计两个子空间的相似度。根据核Fisher准则,基于类间典型差与类内典型差的比率建立核子空间的相关性来得到核典型相关性鉴别分析(KDCC)算法。在ORL、NUST603、FERNT和XM2VTS人脸库上的实验结果表明,该算法能够更有效提取样本特征,在识别率上要优于典型相关性鉴别分析(DCC)和核鉴别转换(KDT)算法。

In this study,we propose a new kernel discriminant for learning and recognition of image sets using canonical correlation. Each image set is mapped into a high-dimensional feature space. The corresponding kernel space is then constructed by a kernel linear discriminant analysis.The similarity of two kernel subspaces is assessed by calculating the canonical difference between them. According to the kernel Fisher discriminant, a Kernel Discriminant Analysis of Canonical Correlation algorithm is derived to establish the correlation between the kernel subspaces based on the ratio of the canonical differences of the between-classes to those of the within-classes. The experimental results on the ORL, NUST603, FERNT and XM2VTS database demonstrate that the proposed method can efficiently extract the features of the images. Moreover, the recognition rate of the proposed algorithm outperforms DCC and KDT.