基于概率隐含语义分析模型的人脸识别算法  

New face recognition algorithm based on probabilistic latent semantic analysis

作　　者：邹修明[1,2] 孙怀江[1] 杨赛[3] Zou Xiuming SUn Huaijiang Yang Sai(School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing 210094, China School of Computer Science and Technology, Huaiyin Normal University, Huaian 223300, China School of Electrical Engineering, Nantong University, Nantong 226019, China)

机构地区：[1]南京理工大学计算机科学与工程学院,江苏南京210094 [2]淮阴师范学院计算机科学与技术学院,江苏淮安223300 [3]南通大学电气工程学院,江苏南通226019

出　　处：《南京理工大学学报》2016年第5期594-598,共5页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(61402192);江苏省高校自然科学基金重大资助项目(15KJA460004)

摘　　要：该文提出一种基于概率隐含语义分析(PLSA)的新的人脸识别算法。首先建立人脸图像的词袋模型,然后使用概率隐含语义分析模型得到词袋特征在隐含主题空间中的分布,并将其作为人脸图像的最终语义特征表示,最后采用支持向量机(SVM)对人脸进行识别。Multi-PIE和人脸识别竞赛(FRGC)数据库上的实验结果表明,该文方法的性能优于目前多个人脸识别方法。A new face recognition algorithm based on the probabilistic latent semantic analysis（ PLSA） is proposed. Firstly the bag-of-word model of the face image is contructed. Then the PLSA model is used to get the distribution of the bag-of-word model in the latent topic space as the final semantic features of the face image. Lastly images are classified and recognized by the support vector machine（ SVM）. Experimental results on the multi-PIE and the face recognition grand challenge（ FRGC） datasets show that the method gets higher classification accuracies than the current methods for the face recognition.

关 键 词：人脸识别 概率隐含语义分析 词袋模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]