计算机工程2012，Vol.38Issue(9)：193-196,4.DOI:10.3969/j.issn.1000-3428.2012.09.058
基于特征融合和流形增强的视频人脸识别
Video Face Recognition Based on Feature Fusion and Manifold Enhancement
摘要
Abstract
A new approach of face recognition in videos, named FFME, is proposed. Faces are divided into different regions. Different local feature descriptors, which are selected against the variation of some region, are combined for face representation. K-NN model is built for feature vectors classification. A sum rule is followed by FFME to combine the individual classification results. Manifold is introduced for constructing a set of reference face images to re-rank the top candidate images, which is approved later to enhance the face recognition rates. Experimental results on the video-based face database Mobo and Honda/UCSD show that FFME is superior to other methods such as principal components analysis, linear discriminant analysis, hidden Markov model, locally linear embedding and manifold-manifold distance.关键词
特征融合/流形/视频人脸识别/局部特征/全局特征Key words
feature fusion/ manifold/ video face recognition/ local feature/ global feature分类
信息技术与安全科学引用本文复制引用
黄存东,刘仁金,杨思春..基于特征融合和流形增强的视频人脸识别[J].计算机工程,2012,38(9):193-196,4.基金项目
安徽省高校教学研究基金资助重点项目(20101689) （20101689）
安徽省自然科学基金资助项目(11040606M150) （11040606M150）
安徽省高校自然科学研究基金资助重点项目(KJ2009A054,KJ2011A048) （KJ2009A054,KJ2011A048）
