A Gender Recognition System Based on Facial Image

Abstract

A gender recognition system (GRS) based on facial images can be embedded in different areas such as surveillance, human-robot interaction, targeted advertising, etc. Traditional facial feature-based recognition systems extract and analyze textures on a face. Although such approaches perform well under certain controlled situations, they may fail due to variations of faces in images, which is very common in real-life images. To overcome such problems, we need to have an effective combination of a feature descriptor, representation, and classifier providing better accuracy. Recently, many recognition problems are tackled by using the Deep Neural Network (DNN) such as Convolutional Neural Network (CNN). However, deep learning needs a large number of images, which is not usually available, to work as expected. We propose a model that combines handcrafted features with CNN to overcome the shortcomings including handling of variations in imaging, such as the illumination and pose variations, and the necessity of voluminous training sets. Experimental results also show that the proposed method performs better than the available gender recognition approaches.