基于分块二维局部保持鉴别分析的二级人脸识别方法

Two-level face recognition method based on modular two-dimensional locality preserving discriminant analysis

在对二维局部保持鉴别分析(2DLPDA)研究的过程中,发现在将样本分块后,对相同位置的块组成的新的样本集独自使用2DLPDA方法,可以有效地将测试样本的类别锁定在一个很小的范围内。由此提出一种基于分块二维局部保持鉴别分析的二级人脸识别方法。在第一阶段首先对样本进行分块,然后独立对相同位置块所组成的新样本集进行2DLPDA,并以此提取出测试样本被锁定的类别范围;之后在该缩小的类别范围内,进行二级人脸识别过程。提出两种方案,一种是二级采用协同表示分类(CRC)算法,另一种是二级采用最近邻分类(NNC)算法来对测试样本的类别进行进一步的识别。在ORL人脸库上的实验结果表明,所提出的方法对于提高识别率有效。

In the research of Two-Dimensional Locality Preserving Discriminant Analysis（ 2DLPDA）,it is found that if samples are divided into blocks and all the blocks of the same position are gathered together to make up a new subset,then2 DLPDA is carried out on the new subset,the test sample can be effectively locked into a very small range of classes. In this paper a two-level face recognition method was presented based on modular two-dimensional locality preserving discriminant analysis. In the first stage,samples were firstly divided into blocks,and then 2DLPDA was carried out on the new subset which made up by the blocks in the same position to seek out the category which the test sample was locked in. In the second stage in this category,the final class of the test sample was recognized. In this paper,two different schemes were put forward for the second stage. The first one was the collaborative representation based classifier algorithm,the other was the nearest neighbor based classifier algorithm. Experimental results on ORL face database show that the proposed method is effective to improve the recognition rate.