中文摘要：

提出一种新的三阶分段光滑函数，构造三阶光滑支持向量机模型（ TPSSVM）。理论证明新三阶分段光滑函数对正号函数的逼近程度更高。在处理多类问题时，提出一种基于编码方式的一对多光滑支持向量机分类方法。对于人脸识别问题，通过主成分分析（ PCA）进行特征提取，并利用多分类光滑支持向量机对人脸特征图像进行训练和测试。应用于ORL人脸库和FERET人脸库的测试结果表明，多分类光滑支持向量机比传统的识别方法有更高的识别率。

英文摘要：

A new three-order piecewise function was used to smoothen the model of Support Vector Machine （ SVM） and a Third-order Piecewise Smooth SVM （ TPSSVM） was proposed. By theory analyzing, approximation accuracy of the smooth function to the plus function is higher than that of the available. When dealing with the multi-class problem, a coding method of multi-class classification based on one-against-rest was proposed. Principal Component Analysis （ PCA） was employed to extract the main features of face image set, and multi-class classification of smooth SVM was used for face recognition. The experimental results on ORL and FERET face databases show that the recognition rate of smooth SVM for multi-class classification is better than the traditional identification methods.