Age-Related Features From Face Image For Hierarchical Age Estimation

Authors

  • Arif Sameh Arif

Abstract

Assigning studied face image to the related age is a challenging task due to the different attributes of aging patterns along different age stages. Different periods of human age are combined by different detailed effects on the human face, which belong to the specific period more than the others. Yet, there are other types of aging continue over all age periods. This paper utilizes the general-age changes and specific-age changes to build a hierarchical technique for age estimation. Face smoothness is adopted for classifying the age into three main age classes, young, adult and senior. For each of them, inner age progression signs are adopted for accurate estimation. In childhood, changes face shape caused by dramatic changes in the size and shape of the head. In adulthood, an adaptive from of Local Binary Pattern (LBP) is proposed as candidate features since this period witness changes in texture nature. For senior faces, gradient features are adopted to describe face lines, folds and wrinkles.  As such features are produced in huge number, Feature Selection Method (FSM) is used to choose efficient features and ignore weak ones. After selecting the best possible set of features, Support Vector Machine (SVM) classifier is used to estimate related ages. Regarding State of Art in age estimation, this paper yielded encouraging results which yielded better results than other published papers in age estimation.

Published

2020-12-01

Issue

Section

Articles