中文摘要：

提出运用多线性核主成分分析（MKPCA）的一种新方法进行掌纹识别。首先MKPCA通过非线性变换,将输入样本图像向高维特征空间F上投影,运用多线性主成分分析（MPCA）直接对掌纹张量进行降维,得到低维的投影张量;然后掌纹图像向张量子空间上投影提取特征向量;最后计算特征向量间的余弦距离进行掌纹匹配。运用PolyU掌纹图像库,对本文算法进行测试。实验结果表明,与传统算法相比,本文算法的识别率（RR）最高为99%,特征提取和匹配总时间为1.832 s,满足实时系统的要求。

英文摘要：

A new palmprint recognition method is proposed based on multilinear kernel principal component analysis（MKPCA）.The MKPCA projected the input samples to high dimensionality feature space F by nonlinear transformation.In the feature space F,the MPCA was first used to operate directly on the tensor objects to obtain the low dimensionality projection tensors.Then the palmprint images were projected onto the tensor subspace for extracting the feature vectors.Finally,the cosine distance between two feature vectors was calculated to match palmprints.The new algorithm was tested in PolyU plmprint database.The experiment results show that compared with the conventional algorithms,the recognition rate（RR） of the new algorithm is the highest up to 99%,and the total time for feature extraction and matching is only 1.832 s,so it meets the real-time system specification.