A new face recognition system based on Kernel maximum between-class margin criterion (KMMC)

Keywords: Face recognition; Feature extraction; Small sample size; Kernel maximum between-class margin criterion Abstract. To avoid small sample problem in pattern recognition, the paper uses KMMC (Kernel maximum between-class margin criterion method) as the basic extraction method for face recognition, which is based on the maximum difference of between-class scatter and within-class scatter in feature space. The objective of KMMC is to seek an optimal set of discriminant vectors as the projection axis to do some projection transformation, and to make the between-class scatter of feature space sample maximum, the within-class scatter minimum, theoretically solved the problems that can not be solved due to singularity of within-class scatter and demonstrates its efficiency of feature extraction furthermore. The test results show the validity of this method on ORL database. At last, it designed and implemented face recognition system based on KMMC by using Matlab7.1.