摘要
本文提出一种新的基于核Fisher判决分析(简称KFDA)的脸谱识别方法。即首先应用KFDA提取脸谱特征,然后,进行脸谱识别。利用标准的AT&T脸谱数据库对KFDA特征提取方法和PCA、FDA以及ICA特征提取方法进行比较,最后使用线性支持向量机(简称SVM)进行分类和识别,实验结果显示基于KFDA特征提取脸谱识别方法的识别率明显优于其它三种脸谱识别方法的识别率。
A new face recognition method is proposed. In this method, Kernel Fisher Discriminant Analysis (KFDA) is combined with Linear Support Vector Machine (SVM). KFDA is a new non-linear technique for extracting features. KFDA-based face recognition method is tested and compared with PCA, FDA and ICA-based face recognition methods using the same publicly available AT&T database. Experiment results indicate that the performance of KFDA-based face recognition method is superior to the others.
出处
《电路与系统学报》
CSCD
2003年第5期57-61,共5页
Journal of Circuits and Systems
关键词
核Fisher判决分析
支持向量机
线性Fisher判决分析
主分量分析
独立分量分析
Kernel Fisher Discriminant Analysis (KFDA)
Support Vector Machine (SVM)
Linear Fisher Discriminant Analysis(FDA)
Principal Component Analysis(PCA)
Independent Component Analysis(ICA)
