File(s) under permanent embargo
Using age information as a soft biometric trait for face image analysis
chapter
posted on 2020-01-01, 00:00 authored by Haoyi Wang, Victor Sanchez, Wanli Ouyang, Chang-Tsun LiChang-Tsun LiSoft biometrics refers to a group of traits that can provide some information about an individual but are inadequate for identification or recognition purposes. Age, as an important soft biometric trait, can be inferred based on the appearance of human faces. However, compared to other facial attributes like race and gender, age is rather subtle due to the underlying conditions of individuals (i.e., their upbringing environment and genes). These uncertainties make age-related face image analysis (including age estimation, age synthesis and age-invariant face recognition) still unsolved. Specifically, age estimation is concerned with inferring the specific age from human face images. Age synthesis is concerned with the rendering of face images with natural ageing or rejuvenating effects. Age-invariant face recognition involves the recognition of the identity of subjects correctly regardless of their age. Recently, thanks to the rapid development of machine learning, especially deep learning, age-related face image analysis has gained much more attention from the research community than ever before. Deep learning based models that deal with age-related face image analysis have also significantly boosted performance compared to models that only use traditional machine learning methods, such as decision trees or boost algorithms. In this chapter, we first introduce the concepts and theory behind the three main areas of age-related face image analysis and how they can be used in practical biometric applications. Then, we analyse the difficulties involved in these applications and summarise the recent progress by reviewing the state-of-the-art methods involving deep learning. Finally, we discuss the future research trends and the issues that are not addressed by existing works. We also discuss the relationship among these three areas and show how solutions within one area can help to tackle issues in the others.
History
Title of book
Deep biometricsSeries
Unsupervised and semi-supervised learningChapter number
1Pagination
1 - 20Publisher
SpringerPlace of publication
Cham, SwitzerlandPublisher DOI
ISSN
2522-848XeISSN
2522-8498ISBN-13
9783030325824Language
engPublication classification
B1 Book chapterExtent
13Editor/Contributor(s)
Richard Jiang, Chang Li, Danny Crookes, Weizhi Meng, Christophe RosenbergerUsage metrics
Categories
No categories selectedKeywords
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC