基于多线性主元分析和FLD的掌纹识别

Palmprint recognition based on multilinear principal component analysis and FLD

为了减少高维对计算成本的影响,同时提取有利于分类的判别特征,提出运用多线性主元分析(MP-CA)与FLD相结合的方法进行掌纹识别。运用MPCA直接对掌纹张量进行降维和特征提取,低维特征向量作为FLD的输入,提取判别特征向量,计算特征向量间的余弦距离进行掌纹匹配。PolyU掌纹图像库的实验结果表明,与主元分析(PCA)、PCA+FLD、二维主元分析(2DPCA)、独立元分析(ICA)和MPCA相比,该算法的识别率(RR)最高为99.91%,特征提取和匹配总时间为0.398s,满足实时系统的要求。

In order to decrease the effect of computational cost due to high dimensionality and extract the discriminant feature vectors which were propitious to classify, this paper proposed a new palmprint recognition method based on muhilinear principal component analysis （MPCA） and FLD. First used the MPCA to operate directly on the original tensor objects. The low dimensional feature vectors were the input of FLD to extract the discriminant feature vectors. Calculated ＇the cosine distance between two feature vectors to match palmprints. The experiment results of PolyU palmprint database show that compared with principal component analysis （PCA） , PCA ＋ FLD, 2-dimension principal component analysis （2DPCA） , independent component analysis （ ICA）, and MPCA, the recognition rate （RR） of the new algorithm is the highest which is 99.91% ; and all the time for feature extraction and matching is 0. 398 s, so it meets the real-time system specification.