| 基于监督式ICA的人脸识别 |
| |
| 引用本文： | 张丽丹,黄凤岗,李先伟. 基于监督式ICA的人脸识别[J]. 哈尔滨工程大学学报, 2006, 27(5): 748-751 |
| |
| 作者姓名： | 张丽丹 黄凤岗 李先伟 |
| |
| 作者单位： | 哈尔滨工程大学,计算机科学与技术学院,黑龙江,哈尔滨,150001;哈尔滨工程大学,计算机科学与技术学院,黑龙江,哈尔滨,150001;哈尔滨工程大学,计算机科学与技术学院,黑龙江,哈尔滨,150001 |
| |
| 摘 要： | 特征提取技术是人脸识别的关键技术,很大程度上决定着识别结果的好坏.而传统的基于独立分量分析的特征提取方法不仅速度慢且未完全利用特征信息,故提出一种监督式特征提取方法.将类信息引入到独立分量的求取过程中,从而得到更有利于分类的独立分量.这里,主要考虑类内散布度对分类的影响,并推导出新的独立分量迭代公式,将其应用到人脸识别问题中.通过在3个标准的人脸数据库上的实验表明,监督式ICA方法在识别率和识别时间上都优于传统的ICA方法.
|
| 关 键 词： | 独立分量分析 人脸识别 最近邻分类器 监督学习 |
| 文章编号： | 1006-7043(2006)05-0748-04 |
| 修稿时间： | 2005-06-27 |
Face recognition based on supervised independent component analysis |
| |
| ZHANG Li-dan,HUANG Feng-gang,LI Xian-wei. Face recognition based on supervised independent component analysis[J]. Journal of Harbin Engineering University, 2006, 27(5): 748-751 |
| |
| Authors: | ZHANG Li-dan HUANG Feng-gang LI Xian-wei |
| |
| Abstract: | Feature extraction is a key factor in face recognition methods and significantly determines recognition performance.To improve upon traditional independent component analysis(ICA),a supervised version of ICA was developed.Class information was introduced into the seeking procedure of independent components,leading to more discriminant independent components.A new iteration formula was developed using scatter degrees within a class.By applying the supervised ICA to three standard face databases,it was concluded that supervised ICA outperforms traditional ICA in both recognition rate and recognition time. |
| |
| Keywords: | independent component analysis(ICA) face recognition nearest neighborhood classifier supervised learning |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
