Kernel Entropy Component Analysis using local mean-based k-nearest centroid neighbour (LMKNCN) as a classifier for face recognition in video surveillance camera systems

In this paper, a new method for face recognition in video surveillance is proposed. Local mean-based k-nearest centroid neighbour (LMKNCN) is a recently proposed method for classifying data which has been proven to be more appropriate than other classifiers such as k-nearest neighbour (KNN), K-Nearest Centroid Neighbour (KNCN), and local mean-based k-nearest neighbour (LMKNN). Kernel Entropy Component Analysis is a new extension of 1-D PCA-based feature extractions methods enhancing the performance of PCA-based methods. In the proposed method in this paper, LMKNCN is used as a classifier in KPCA method. Moreover, the Extensive experiments on surveillance camera faces database (SCfaces) and Head Pose Image database reveal the significance of the proposed method.