通用学习框架在单样本人脸识别中的应用研究

RESEARCH AND APPLICATION OF GENERIC LEARNING FRAME IN FACE RECOGNITION WITH SINGLE SAMPLE

针对单样本人脸识别问题,提出一种基于通用学习框架的人脸识别方法。以大量的通用样本与各个单样本按一定比例叠加的方式,增加每个类的训练样本总数,有效地运用了2DPCA方法进行特征抽取,将所有样本投影到特征子空间,再根据最大隶属度原则完成人脸识别,明显提高了识别率。该方法的有效性分别在ORL及FERET人脸数据库上得到了验证。

Aiming at the problem of face recognition with single sample,we propose in this paper a recognition method that uses generic learning frame. It increases the total number of training samples of each class in the way of piling up in certain proportion the large amount of generic samples and each single sample,effectively employs the 2DPCA method for features extraction. All the samples are projected onto the feature subspace,and the face recognition is accomplished according to maximum membership principle,thus the recognition rate is remarkably raised. The effectiveness of the proposed method has been verified on ORL and FERET face database.