基于自调进度稀疏表示的人脸识别算法研究

项目名称： 基于自调进度稀疏表示的人脸识别算法研究

项目编号： No.61501230

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 无线电电子学、电信技术

项目作者： 朱旗

作者单位： 南京航空航天大学

项目金额： 21万元

中文摘要： 近年来，稀疏表示是人脸识别领域中的一个研究热点。相比传统的稀疏表示模型，基于Lp范数（0

中文关键词： 图像识别；稀疏表示；人脸识别

英文摘要： In recent years, sparse representation method is a hot research topic in the field of face recognition. Compared with the traditional sparse representation model, the Lp-norm (0<p<1) based model is better, but it needs to solve the non-convex optimization problem, which is hard to compute and an algorithm may easily get trapped in the local minima. Aiming at solving the above problem of the Lp-norm based sparse representation model, we will study the following issues: (1) We will apply the self-paced learning idea to solving Lp-norm based sparse representation model, which can reduce the influence of noise and outliers in the learning process, and avoid getting trapped in the poor local minima. (2) In order to handle large-scale face image data, we will design the self-paced learning based nonlinear sparse representation based classification method (SRC). (3) We will also design the fast algorithm for self-paced learning based SRC, and try to project the samples onto the low dimensional space, in which the self-paced learning based SRC can obtain higher classification accuracy than that in original space. (4) We seek to preserve the local geometric structure of the data and enhance the discriminant ability of the data in designing the sparse regression model. We also introduce the self-paced learning idea to this model to obtain more robust and sparse solution, which can be expected to have better face image representation and classification result.

英文关键词： Image Classification;Sparse Representation;Face Recognition