2D—FDA和ProximalSVM在人脸识别上的联合应用

2D-FDA Plus PSVM Using in Face Recognition

对人脸图像进行二维Fisher鉴别分析（2D—FDA）的特称抽取与最临近支持向量机（ProximalSVM）的分类进行组合。首先把人脸图像按测试样本和训练样本进行划分。对训练样本进行2D—FDA特征抽取，得到抽取不同特征数目的具有最大鉴别信息的特征向量。然后再把此特征向量与测试样本相结合，用最简单的支持向量机进行分类，得到比用最小欧氏距离方法更高的识别效率，从而说明这两种方法的组合在人脸识别应用中发挥了各自的优点。

The paper gives a method that combined Two Dimensional Fisher Discrimant analysis （2D-FDA） and Proximal SVM in face recognition. Firstly we divide the train sample from the whole, and the rest is the test sample. Then we extract most useful features using 2D-FDA method. After that, we project both train sample and test sample into the extractive feature space. Finally we classify the test sample used the Proxmal SVM method. The higher recogniation rate shows that the combined method has a better classification performances.