Face Recognition Using Pearson Correlation and HOG with Single Training Image Per Person

Abstract

Face recognition with training single sample per person (SSPP) is a full of challenging task because in such a scene it is difficult to predict the face correct only by seeing some one one time. Considering the fact that different portions of human faces have different information to face recognition, we use histograms of oriented gradients (HOG) algorithm to compute the features based on framework of face matrix and use Pearson correlation to classify these features with SSPP. HOG extracts the most framework information both training sample and testing sample, Pearson correlation detects the same information between them. This method which includes HOG and Pearson correlation makes well difference in variation representation, showing a good performance in face recognition with SSPP.