基于不同Margin的人脸特征选择及识别方法

李伟红; 陈伟民; 杨利平; 龚卫国

doi:10.3724/SP.J.1146.2005.01567

基于不同Margin的人脸特征选择及识别方法

doi: 10.3724/SP.J.1146.2005.01567

基金项目:

国家教育部科学研究重点项目(02057)，重庆市自然科学基金重点研究项目(CSTC2005BA2002)和重庆市自然科学基金(CSTC2005 BB2181)资助课题

计量
- 文章访问数: 3428
- HTML全文浏览量: 73
- PDF下载量: 983
- 被引次数: 0
出版历程
- 收稿日期: 2005-12-05
- 修回日期: 2006-04-03
- 刊出日期: 2007-07-19

Face Feature Selection and Recognition Based on Different Types of Margin

摘要

摘要: Margin在机器学习中具有很重要的意义，基于margin的特征选择方法就是从分类的角度对特征集各特征的权重进行分析。该文对不同的margin进行了分析，提出将sample-margin和hypothesis-margin分别作为特征选择标准对SBS特征选择方法进行改进，然后设计具有最佳超参数的SVM多项式分类器进行人脸识别。实验在FRERT人脸图像库上进行并与Relief特征选择方法进行了比较，对SVM和NN分类器的实验结果也进行了分析。实验结果显示：该文提出的人脸识别特征选择及识别方法是有效、适用的。
- 人脸识别;Margin;特征选择;支持向量机(SVM);顺序后退法(SBS)
Abstract: Margin plays an important role in research of machine learning. Margin-based feature selection methods choose the weights of features from the view of classification. This paper analyzes different types of margin and proposed methods to improve the Sequential Backward Selection (SBS) method respectively using sample-margin and hypothesis-margin as feature selection criterion. A SVM polynomial classifier, which has optimal hyper-parameters, is then designed for face recognition. Experiments are conducted on FERET face database. Recognition accuracies between the proposed methods and relief feature selection method are compared. Experiments are also conducted by respectively using SVM and Nearest Neighbor (NN) classifier. Experimental results indicate that the proposed feature selection and recognition methods are efficient for face recognition.

HTML全文

参考文献(1)

Langley P. Selection of relevant features in machine learning. In: Proc. AAAI Fall Symposium on Relevance, New Orleans, Louisiana, 1994: 140-144.[2]Gilad-Bachrach R.[J].Navot A, and Tishby N. Margin based feature selection - theory and algorithms. In proceedings of the 21st International Conference on Machine Learning (ICML), Banff, Alberta, Canada, July 4-.2004,:-[3]Crammer K, Gilad-Bachrach R, Navot A, and Tishby N. Margin analysis of the lvq algorithm. Proc.17th Conference on Neural Information Processing Systems, Banff, Alberta, Canada, 2002: 462-469.[4]Vapnik V. The Nature of Statistical Learning Theory. Spring- Verlag, 1995.[5]Freund Y and Schapire R E. A decision-theoretic generalization of on-line learning and an application to boosting[J].Journal of Computer and System Sciences.1997, 55(1):119-139[6]Buckingham L and Geva L. Lvq is a maximum margin algorithm. In Pacific Knowledge Acquisition Workshop PKAW2000, Sydney, Australia, 2000.[7]Weston J, Mukherjee S, Chapelle O, Pontil M, Poggio T, and Vapnik V. Feature selection for SVMs. In Advances in Neural Information Processing Systems. MIT Press, 2000, (13): 668-674.[8]Grandvalet Y and Canu S. Adaptive scaling for feature selection in svms. In Advances in Neural Information Processing Systems, MIT Press, 2003, (15): 553-560.[9]Guyon I, Weston J, Barnhill S, and Vapnik V. Gene selection for cancer classification using support vector machines[J].Machine Learning.2002, 46(1/3):389-422[10]Ahmad A and Dey L. A feature selection technique for classificatory analysis[J].Pattern Recognition Letters.2005, 26(1):43-56[11]Weihong L and Weiguo G. Feature selection based KPCA, SVM and GSFS for face recognition. Proceeding of 3rd International Conference on Advances in Pattern Recognition, London, 2005, 2: 344-350.[12]Bartlett P. The sample complexity of pattern classification with neural networks: the size of the weights is more important than the size of the network[J].IEEE Trans. on Information Theory.1998, 44(2):525-536[13]Kira K and Rendell L A. The feature selection problem: traditional methods and a new algorithm. In: Proc. of the Ninth National Conf. on Artificial Intelligence, Menlo Park, CA, USA, 1992: 129-134.[14]Nastar C and Ayache N. Frequency-based non-rigid motion analysis[J].IEEE Trans. on Pattern Anal. Machine Intell.1996, 18(11):1067-1079

施引文献

资源附件(0)

访问统计