Zernike moments and machine learning based gender classification using facial images

Abstract

This paper proposes a zernike moments based gender classification system using facial image. Gender classification is achieved by training the Bayesian, Support vector machine, Linear discriminant analysis and Neural network classifiers. The proposed method was evaluated using ORL and faces94 databases. The simulation results indicate the effectiveness of the method in achieving the greater accuracy even under the variations in pose, scale, rotation and occlusion. In particular the neural network classifier was excellent in providing classification accuracy of 95% and 100% respectively with 25 zernike moments for ORL and faces94 database.