Eigenface and SIFT For Gender Classification

Abstract

In this paper, we study and analyze five different approaches for gender classification (Male or Female). The purpose of approaches is to extract the main features from face image and for each approach; we used these features as input to the Support Vector Machine SVM classifier for classification process. In the first approach, we implemented the Principal Component Analysis PCA features as input to the SVM classifier. In the second approach, the resulted parameters of Scale Invariant Features Transform approach are used as input to SVM classifier. In the third approach, we implemented Eigenfaces as input to SIFT, and then the results of SIFT are used as input to SVM classifier also. In the fourth approach, we implemented Eigenfaces to be input to Volume-SIFT (VSIFT) and then used as input to SVM classifier. The last, we modified the VSIFT approach and use as input to SVM classifier. The practical implementation results show that the proposed approach (modified VSIFT) gave us high performance of gender classification than other approaches.The training and testing images set consist of number of faces which are obtained from (Computer Vision Science Research Projects).The aim of this paper is to analyze and discuss the results of applying the eigenfaces into the new proposed approach of modifying Volume-SIFT for classifying the faces into its gender.