基于多源数据融合的人脸识别研究  

Research on Multi-source Data Fusion for Face Recognition

作　　者：叶继华[1] 陈亚慧[1] 胡蕾[1] 

机构地区：[1]江西师范大学计算机信息工程学院,南昌330022

出　　处：《系统仿真学报》2014年第7期1554-1558,共5页Journal of System Simulation

基　　金：国家自然科学基金项目(61262036);江西省教育厅项目(GJJ13228);江西省分布计算工程技术研究中心项目(2012006)

摘　　要：多源人脸信息的融合识别是多源信息融合领域与人脸识别领域的交汇点,它能减少人脸多姿态、表情等因素对人脸识别的影响。通过对对已有的多源人脸的融合技术以及单源人脸的识别技术的分析,提出结合2DPCA、二维Gabor小波和多子空间分析进行多源人脸信息的融合实现多姿态人脸识别。首先,对样本图像进行二维Gabor小波特征描述,然后利用2DPCA降维,最后利用多子空间分析进行多源信息融合完成多姿态人脸的识别。试验证明方法提高了多姿态人脸识别的精度。The fusion recognition of multi-source face information was the intersection of multi-source information fusion area and face recognition area, and it could reduce the influence of face pose and face expression. A method for multi-pose face recognition, which was the combination of 2DPCA, 2D gabor wavelet and subspaces analysis for multi-source information fusion, was provided by the analysis of existing multi-source information fusion technology and monophyletic face recognition technology. First of all, sample images were given feature descriptions by 2D gabor wavelet. Then these feature descriptions were given multi-source information fusion by 2DPCA. At the end, multi-pose face recognition was implemented by using multi-source information fusion based on the subspaces analysis. The experiment results show that the method applied to the multi-pose face recognition improves the accuracy of multi-pose face recognition.

关 键 词：人脸识别 多源信息融合 二维Gabor小波 2DPCA 子空间分析 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]