Abstract: Principal Component Analysis（PCA） or Kernel Principal Component Analysis（KPCA） can only extract the linear or nonlinear features of palmprint, and single classifier recognition rate is very low, this paper proposes a two level classifier for palmprint recognition based on subspace features. Firstly, the PCA and KPCA are used to extract the linear or nonlinear features of palmprint, respectively, and the best fusion coefficient can be calculated by making the total distance of between-classes largest to get the optimal features of palmprint image, the Euclidean distance metric method is used to recognize palmprint image, if the palmprint image category is clearly, the recognition result is obtained, otherwise the palmprint image is put into support vector machine to recognize. Polyu palmprint image library is used to test the performance, the results show that, compared with other palmprint recognition methods, the proposed method has improved the palmprint recognition rate and recognition speed, and false accept rate and false reject rate are reduced.

Key words: palmprint recognition, kernel principal component analysis, Euclidean distance, support vector machine, feature extraction

摘要： 针对单一PCA或PCA只能提取掌纹的线性或非线性特征，单一分类器的掌纹识别率低缺陷，提出一种子空间特征融合的两级掌纹识别方法（PCA-KPCA-SVM）。首先采用子空间特征提取方法PCA、KPCA分别提取掌纹图像线性和非线性特征，然后基于融合特征总类间距离最大准则，计算出最佳的融合系数，得到PCA、KPCA的融合掌纹特征，最后将融合特征输入到欧式距离分类器进行掌纹识别，如果拒绝识别，则输入支持向量机进行二次识别。采用Polyu掌纹图像库进行测试实验，结果表明，相对于对比算法，PCA-KPCA-SVM提高了掌纹识别率，有效降低了掌纹的误识率和拒识率。

关键词: 掌纹识别, 核主成分分析, 欧式距离, 支持向量机, 特征提取