Automatic human age estimation from face images using MLP and RBF neural network algorithms in secure communication networks

Abstract

Age estimation finds application in several contexts including biometric authentication, surveillance, forensic investigations, and the entertainment industries, among others, making it a realistic, complex, and relevant problem in the subfield of machine vision and pattern recognition. This article proposes a system that can determine the age of people by applying the multilayer perceptron neural network technique, feature fusion, and integration. It is imperative to define three fundamental stages of the proposed procedure. One of the methods necessary at the first stage involves face parts detection and resizing. The second activity is feature extraction of the facial regions of the video frames. To do this, we select and use features through the use of Gabor Filters, SIFT, and LBP with optimal values being picked through combinations. The third step in the process entails making an age range prediction of the face image using neural network algorithms such as MLP and RBF for securing communication networks. For further reduction of dimensions and for getting rid of any possibly overlapping features, independent component analysis (ICA) is used. The datasets adopted for this research work are FG-NET and PAL which are widely acclaimed research data sets. Simply based on the features adopted, the proposed age estimate procedure was seemingly superior to the MLP algorithm and possessed high accuracy for a given range of ages.