人脸识别中基于线性判别分析的边信息方法

在非限制环境下，人脸识别数据库上评估时由于无法获得所有对象完整的类标签信息，只能得到弱标签信息，使得许多人脸识别算法将无法工作。为了解决此问题，本文提出了一种基于线性判别分析的边信息方法，即加权SILD( Side-Information based Linear Discriminant)方法WSILD，只利用边信息就可以很好地工作，其中，类内和类间散布矩阵直接利用边信息计算出来。在美国国防部人脸识别技术FERET及耶鲁大学人脸数据库Yale上的实验支撑了算法的理论分析。当本文提出的WSILD方法使用了多种特征时，与最先进的方法相比，也取得了更好的人脸识别效果。

Research of Side-Information based on Linear discriminate analysis in Face Recognition

In the unconstrained environment, some methods have been evaluated on popular face databases in face recognition, and sometimes we cannot obtain the full class label information of all the subjects instead of the weak label information, many multi-class methods fail to work due to the lack of full class label information. To effectively utilize the side-information in such case, a Side-Information based Linear Discriminate Analysis (SILD) was proposed in this paper, in which the within-class and between-class scatter matrices were directly calculated by using the side-information. Moreover, we theoretically prove that our SILD method is equivalent to FLDA when the full class label information is available. Experiments on Face Recognition Technology (FERET) and Yale databases show the better recognition efficiency comparing with the latest methods. Compared with the best algorithms, a good face recognition effect was also obtained using multi-characters in WSILD.