一种基于E-HMM的选择性集成人脸识别算法  

A Face Recognition Algorithm Based on Selective Ensemble of E-HMMs

作　　者：李金秀[1] 高新波[1] 杨越[1] 肖冰[1] 

机构地区：[1]西安电子科教大学电子工程学院,西安710071

出　　处：《电子与信息学报》2009年第2期288-292,共5页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(60202004);教育部长江学者和创新团队支持计划;新世纪优秀人才支持计划(NCET-04-0948)资助课题

摘　　要：基于嵌入式隐马尔可夫模型(Embedded Hidden Markov Model,E-HMM)的人脸识别方法的识别性能依赖于模型参数的合理选择。提出了一种基于E-HMM的多模型选择性集成人脸识别算法,选择出个体精度高且互补性强的模型来进行集成的人脸识别。实验结果表明,与传统的基于E-HMM的人脸识别方法相比,新算法不仅可以获得更好、更稳定的识别效果,而且具有更强的泛化能力。The performance of Embedded Hidden Markov Model （E-HMM） based face recognition algorithm heavily depends on the selection of model parameters. A selective ensemble of multi E-HMMs based face recognition algorithm is proposed, selecting many Comparing with the traditional E-HMM based face the proposed method can not only obtain better generalization ability. accurate and diverse models for ensemble face recognition. recognition algorithm, the experimental results illustrate that and more stable recognition effect, but also achieve higher

关 键 词：人脸识别 嵌入式隐马尔可夫模型 模型选择 选择性集成 泛化能力 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]