中文摘要：

SVM是人脸识别中最常使用的一种机器学习领域算法,它通过距离概念得到对数据分布的结构化描述,降低了对数据规模的要求,适合处理人脸图像这种小样本训练集的分类问题。其中SVM的核函数的选择对分类精度影响很大,全局核函数的预测函数对输出进行正确预测的能力较高,而局部核函数具有较强的学习能力,兼顾两者特点,使用结合RBF核和Sigmoid核的混合核来设计SVM分类器进行识别。针对ORL库进行PCA特征提取,然后使用基于混合核的SVM分类器进行识别分类。实验结果表明,在识别率上,基于该混合核函数的SVM分类器比基于普通核函数SVM分类器要更占优势。

英文摘要：

Support vector machine（ SVM） is one of the most commonly used algorithm in machine learning when it comes to face recognition,it gets structured description of data distribution by the conception of distance and reducing the requirements of data volume,so it＇s very suitable for the face recognition of small sample of the training set. The selection of kernel function of SVM has a great influence on the classification accuracy,global kernel function has the strong ability of generalization but weak in learning,local kernel function is the opposite,taking into account of both advantages,SVM classifier is designed by using the mixture of RBF core and Sigmoid core for identification. using PCA algorithm to extract feature ORL face database firstly,and then using combined-kernel function of SVM classifer to do classification. The result proved that combined-kernel function of SVM has higher recognition rate than traditional single kernel function.