Image-based age-group classification design using facial features

Abstract

Image recognition of a user plays an important role in designing intelligent and interactive behaviors for a domestic or service robot. In this paper, an image-based age-group classification method is proposed to estimate three levels of age groups, namely child, adult and the elderly. After face detection from the acquired image frame, human facial area is extracted and 52 feature points are located by using Lucas-Kanade image alignment method. These feature points and corresponding located facial area are used to build an active appearance model (AAM). After facial image warping, the texture features are sent to a support vector machine (SVM) to estimate the level of age group. In the experimental results, the average recognition rate of the proposed method is 87%. It will improve the interaction capability of robot in a friendly manner.