Soft Biometrics from Face Images Using Support Vector Machines

Abstract

Soft biometrics, such as age, gender, and ethnicity, are useful for many applications in practice. For instance, in business intelligence, it is helpful to automatically extract and compute the statistics of potential customers, such as the number of males and females; the number of young, adult, and senior people; or the number of Caucasian, African American, or Asian people. It is also helpful to use soft biometrics to improve the performance of traditional biometrics for human identification, such as face recognition. Different methods can be developed to recognize the soft biometric characteristics from face images. In this chapter, we present the application of the support vector machines (SVM) to learn an estimator or recognizer to extract these soft biometrics. We will mainly focus on age estimation, while the gender and ethnicity classification will also be discussed. Both classification and regression will be considered. The combination of regression and classifiers based on the SVM will also be described which is useful especially for age estimation.