Application of Face Recognition with Graph Embedding Kernelization

At present, human face technology is applied in many fields. The most important factor to enhance recognition ability is to build a model that can maximize inter-class diversity as well as minimizing intra-class compactness. In this aspect, traditional methods which are Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) have some unresolved problems such as data overlapping. So Kernel Discriminant Embedding (KDE) was introduced. KDE includes three mechanisms which are Kernel trick, Graph Embedding (GE) and Fisher´s criterion (FC), so it can capture face data character efficiently. The process of face recognition by KDE method was presented, superiority and cost of time were also mentioned after evaluated by FRGC database.