B2DPCA和ELM人脸识别算法研究  

Face Recognition Algorithm Based on B2DPCA and ELM

作　　者：李定珍[1] 郭建昌[1] 

机构地区：[1]南阳理工学院电子与电气工程学院,河南南阳473004

出　　处：《液晶与显示》2013年第3期440-445,共6页Chinese Journal of Liquid Crystals and Displays

基　　金：河南省教育厅科技攻关项目(No.12B510024)

摘　　要：提出一种新型、高效的基于B2DPCA(双向二维主成分分析)和ELM(极端学习机)的人脸识别算法,该算法是根据曲波变换分解人脸图像和一种改进的降维技术,通过B2DPCA生成识别特征集来训练和测试ELM分类器,提高识别精度。通过大量实验,并把实验结果与现存技术进行比较,结果表明B2DPCA+ELM算法有效地提高了识别准确率,并降低了对原型数量的依赖。将来有望能把局部特征和基于曲波分解的全局信息结合起来应用到识别精度和分类速度上。A new human face recognition algorithm was proposed based on B2DPCA（bidirectional two dimensional principal component analysis） and ELM（extreme learning machine）.This method was based on curvelet image decomposition of human faces and an improved dimensionality reduction technique.Discriminative feature sets were generated by using B2DPCA to train and test ELM classifier.The recognition accuracy can be improved by using this method.Extensive experiments were performed by using databases and results were compared with state of the existing techniques.The results showed recognition accuracy and minimal dependence on the number of prototypes were significantly improved by using B2DPCA and ELM algorithm.The local characteristics and global information based on curvelet decomposition are expected to apply to the recognition accuracy and speed of classification in the future.

关 键 词：人脸识别 双向二维主成分分析 极端学习机 降维技术 识别准确率 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]