一种新颖的核鉴别通用矢量集算法  

A Novel Kernel Discriminative Common Vectors Algorithm

作　　者：林玉娥 顾国昌[1] 刘海波[1] 沈晶[1] 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,哈尔滨150001

出　　处：《计算机辅助设计与图形学学报》2008年第12期1636-1640,共5页Journal of Computer-Aided Design & Computer Graphics

基　　金：中国博士后基金(20060400809);黑龙江省青年科学技术专项基金(QC06C022)

摘　　要：为了进一步增强鉴别通用矢量集算法的性能,提出一种核鉴别通用矢量集算法.首先利用核函数将原始样本隐式地映射到高维特征空间;然后在高维特征空间里利用再生核理论建立鉴别通用矢量集算法的等价求解模型;最后根据新的求解模型,应用二次Gram-Schmidt正交化方法求出核类内零空间中的鉴别矢量集.在人脸库上的实验结果验证了文中算法的有效性.In order to improve discriminant ability of the discriminative common vectors （DCV） algorithm, a novel kernel discriminative common vectors （KDCV） algorithm is developed. The kernel function is used firstly to project the original samples into an implicit space called feature space by nonlinear kernel mapping, then the equivalent solution model of the KDCV algorithm are established by the theory of reproducing kernel in the feature space. Finally, according to the new solution model of the KDCV algorithm, the discriminant vectors in the null space of the kernel within-scatter matrix are extracted by performing the Gram-Schmidt orthogonalization twice. Experimental results on face database demonstrate the effectiveness of the proposed method.

关 键 词：核函数 特征空间 核类内零空间 鉴别矢量集 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]