一种组合类别信息的非负矩阵分解方法及其应用

Method of Class-information-incorporated Non-negative Matrix Factorization and Its Application

基于非负矩阵分解理论,提出一种新的有监督的特征提取方法,它具有二个特点:一是在特征提取过程中它直接利用训练样本的类别信息,二是在计算上仍然采用与非负矩阵分解方法相同数学公式,因此这种新特征提取方法被称为组合类别信息的非负矩阵分解(CINMF)方法。另外,在分类时本文提出了基于两种特征融合的分类策略进一步提高CINMF方法的识别率。通过在YALE人脸库和ORL人脸库上进行实验,结果表明本文提出的新方法在识别率方面整体上好于原非负矩阵分解方法,甚至超过常用的主成分分析法(PCA)。

A novel supervised feature extraction method based on non-negative matrix factorization （NMF） was proposed. The new method has two traits： one is to sufficiently utilize a given class label of training sample in feature extraction and the other is to still follow the same mathematical formulation as NMF, so the new feature extraction method is named class-information-incorporated non-negative matrix factorization （C1NMF）. Besides, in order to further improve recognition rate, a new classification strategy was proposed based on fusion of two kinds of feature vector. The experimental results on YALE face database and ORL face database show that the new method is better than original NMF in terms of recognition rate, and even outperform PCA.