Dispersion Measures Describe Face Smoothness for Gender and Age Recognition

Abstract

Face biometrics play a crucial role in describing gender and age changes. Male and female faces differ in terms of smoothness due to the formative properties, nature of lifestyle, and use of cosmetics. For age progression, faces record different levels of face vitality-from smooth face within childhood to wrinkles and folds in senior adult faces. This paper presents a set of dispersion measures to represent face features over gender and age variation. Besides, we inspect the exchanging effects between gender differences and life stages on the yielded results. These dispersion features are represented using localized formulas of statistical measures, such as Standard Deviation (SD), Statistical Ratio (SR) and Roughness Value (RV). Such statistical measures inspect the differences of image values regarding specific reference. The SD inspects the differences between a set of values and their average, while SR gauges the SD for set of values normalized by their average. The RV measure inspects the difference between the successive values normalized by their differences from the average. This paper adopts the localized forms of these measures to avoid the effects of the challenges (illumination and rotation) that dataset images suffer from. These adopted features provide significant discriminating powers over different ages and genders. The yielded results show that the adopted measures recorded mutual effects between age estimation and gender recognition.