鉴别投影嵌入及其在人脸识别中的应用

Discriminant Projection Embedding with Its Application to Face Recognition

该文提出了一种新的监督线性降维方法,称为鉴别投影嵌入(Discriminant Projection Embedding,DPE)。和常用的线性鉴别分析相比,鉴别投影嵌入可以更好地保留类内的局部几何位置信息和提取类间的鉴别结构信息。在人脸识别公用数据库上进行了一系列的实验,实验结果表明了该文方法的可行性和有效性。

A new supervised linear dimensionality reduction method called Discriminant Projection Embedding （DPE） is proposed. Compared with widely-used Linear Discriminant Analysis （LDA）, DPE can preserve the within-class neighboring geometry and extract between-class relevant structures for classification more efficient. Experimental results on public face databases show the feasibility and efficiency of DPE.