Abstract: Aiming at the limitation of ignoring the local structure feature of the traditional linear dimensionality reduction methods，a new semi-supervised manifold learning is proposed.On the basis of the character of the face image，this method gets K-nearest neighbors of each sample by calculating the image euclidean distance，and the adjacency matrix of unsupervised discriminant projection is modified accordingly.Finally，the proposed method that combines labeled samples with modified unsupervised discriminant projection is presented to achieve optimal geometric projection.Extensive experimental results on several public face databases validate the correctness and effectiveness of the proposed approach.

Key words: unsupervised discriminant projection, nonparametric discriminant analysis, image euclidean distance, manifold learning, semi-supervised learning, face recognition

摘要： 针对传统线性降维方法忽略数据局部结构特性的问题，提出了一种基于半监督流形学习的方法。针对人脸识别采用图像欧式距离来选择各样本点的K近邻，由此得到修改后无监督判别投影中的邻接矩阵，在传统的无监督判别投影中，融入类标签信息获得几何最优投影。通过在人脸库上的大量比较实验，验证了该方法的准确性和有效性。

关键词: 无监督判别投影, 非参数鉴别分析, 图像欧式距离, 流形学习, 半监督学习, 人脸识别