融合PCA与LDA变换的仿生人脸识别研究  

Biomimetic pattern face recognition integration of PCA and LDA transform

作　　者：蒋加伏[1] 袁承伟[1] 

机构地区：[1]长沙理工大学计算机与通信工程学院,长沙410004

出　　处：《计算机工程与应用》2010年第19期160-163,共4页Computer Engineering and Applications

基　　金：湖南省自然科学基金(No.06jj50109);湖南省科技计划项目(No.06fj3161)~~

摘　　要：就基于PCA与LDA变换的传统人脸识别方法识别率低但特征提取过程中维数低和基于K-L变换的仿生人脸识别方法识别率高但在特征提取过程中维数过高的的问题,将两者的优点相结合,提出了一种基于PCA与LDA变换的仿生人脸识别新方法。通过PCA与LDA变换对训练人脸样本进行特征提取,然后构建各类样本的覆盖区域。再通过判断待识别人脸特征在各覆盖区域的归属情况来识别人脸。实验收到了预期的效果,证明了方法的可行性。A new method of biomimetic pattern face recognition theory based on PCA and LDA transform is proposed.This method has solved the low recognition rate and the excessively high dimension problem,the features of human face on the training samples are extracted through PCA and LDA,and are used to construct the cover region of each kind of sample.The person face is distinguished through the judgment that the person face characteristic belongs to which kind of cover region or doesn＇t not belong to any region.The experiment has received the anticipated effect,and has proven this method feasibility.

关 键 词：主成分分析(PCA) 线性鉴别分析(LDA) K-L变换 仿生模式识别 高维空间几何形体 同源连续性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]