应用于人脸识别的改进局部保持投影算法

Improved Locality Preserving Projections for Face Recognition

摘要:

局部保持投影算法（locality preserving projections,LPP）作为降维算法，在机器学习和模式识别中有着广泛应用。在识别分类中，为了更好的利用类别信息，在保持样本点的局部特征外，有效地从高维数据中提取出低维的人脸图像信息并提高人脸图像的识别率和识别速度，使分类达到一定优化，基于LPP算法结合流形学习思想，通过构造一种吸引向量的方法提出一种改进的局部保持投影算法（reformation locality preserve projections ,RLPP）。将数据集利用极端学习机分类器进行分类后，在标准人脸数据库上的进行试验，实验结果证明，改进后算法的识别率优于LPP算法、局部保持平均邻域边际最大化算法和鲁棒线性降维算法，具有较强的泛化能力和较高的识别率。

Abstract:

As a dimensionality reduction algorithm, Locality Preserving Projections (LPP) is widely used in machine learning and pattern recognition. In the recognition classification, in order to make better use of the category information, in addition to maintaining the local features of the sample points, the low-dimensional face image information is effectively extracted from the high-dimensional data and the recognition rate and recognition speed of the face image are improved. The classification is optimized to a certain extent. Based on the LPP algorithm combined with the manifold learning idea, a reformation locality preserve projections algorithm (RLPP) is proposed by constructing an attractive vector. After the data set is classified by the extreme learning machine classifier, the experiment is carried out on the standard face database. The experimental results show that the improved algorithm has better recognition rate than the LPP algorithm, the local-preserving average neighborhood margin maximization algorithm and the robustness linear dimensionality reduction algorithm，and has strong generalization ability and high recognition rate.