基于集成学习的规范化LDA人脸识别

Regularized Linear Discriminant Analysis for Face Recognition Based on Ensemble Learning

针对人脸识别问题中经常面临的"小样本"问题,在规范化的LDA算法的基础上加以改进,并结合集成学习的方法,利用Adaboost算法,在每一次的迭代过程中引进一个加权函数对难以分离的样本增加权重。增加分类器之间的差异度,从而提高样本在新的特征空间中的可分离性,将识别率提高至98.5%。通过ORL数据库的大量实验表明,该算法比传统算法有更好的性能。

For dealing with the ＂Small Sample Size（SSS）＂ problem in face recognition tasks,regularized Linear Discriminant Analysis（LDA） based on ensemble learning is proposed.By using Adaboost method,weighting function is introduced into the samples,which is closer in the output space in each iteration.So that the separability between these classes is enhanced in the new feature subspace,and the recognition rate is also improved to 98.5%.Experimental results on facial database of ORL show that this method achieves better performance than traditional methods do.