一种应用于人脸识别的非线性降维方法

A novel nonlinear dimensionality reduction approach for face recognition

局部线性嵌入算法(locally linear embedding,LLE)作为一种新的非线性维数约减算法,在高维数据可视化方面获得了成功的应用。然而LLE算法获取的特征从分类角度而言并非最优,而且LLE算法难以获取新样本点的低维投影。为解决这两个缺陷,提出了将非线性的LLE算法和线性判别分析算法(linear discriminant analysis,LDA)相结合的一种新的非线性降维方法,通过ORL、Havard和CMU PIE三个人脸库的实验,结果表明,该方法能够大幅度提高识别率,对光照、姿态及表情变化具有一定的鲁棒性。

Locally linear embedding （LLE） is one of the recently proposed manifold learning algorithms for nonlinear dimensionality reduction, which has demonstrated promising results in visualizing high dimensional data. However, the LLE lacks a parametric mapping between the observation and the low-dimensional output, In addition, since it is developed based on minimizing the reconstruction error, it may not be optimal from classification viewpoint. In this paper, we present a novel nonlinear dimensionality reduction approach for face recognition by fusion of LLE and LDA, Experiments on three public available face databases show the advantages of our proposed novel annroach