Compressive Sensing based Gender Recognition

Abstract

This paper explores an integration of compressive sensing, curvelet transform, and Principal Component Analysis to develop a robust gender recognition method from face images. Compressive measurements of face images leading to a significant reduction in feature space. Here curvelet transform has been used to represent the face images with prominent edges, curvatures, boundaries and to offer sparse representation to apply compressive measurements on detailed subband. To extract the feature vector, Principal Component Analysis is applied on the reconstructed detailed subband. Performance of the proposed method is evaluated by employing different classifiers. The proposed method efficiently handles the effect of Gaussian noise maintaining high accuracy on gender recognition. Extensive experiments on FERET database, is conducted to substantiate our claim.