Soft Biometrics Estimation Using Shearlet and Waveatom Transforms With Three Different Classifiers

Abstract

The goal is to find the best feature extraction, which performs the smallest feature vector length and gives the highest performance. In this paper, we proposed a methodology to extract effective features from facial images using two multiresolution transforms; waveatom and shearlet, for estimating gender, ethnicity, facial expression and age. Three classifiers used to perform the final estimation, which are: Artificial Neural Network (ANN), Support vector machine (SVM) and Self-Organization Map (SOM). A comparative study is made to determine the best extractor and classifier. Experiments carried out on a large database collected from three different databases: US Adult Faces, Extended Cohn-Kanade and FG-NET database. The experimental results of the proposed methodology using waveatom transform proved to be effective in the three classifiers, In contrast of shearlet transform.