Multiple classifiers for age prediction against AAM and ASM

Abstract

In recent years, the researchers on age prediction relied on face pictures to get more attention, due to their important applications in security control and human computer interaction. Age prediction incorporates two processes: traits elicitation and prediction of machine learning. In the aspect of face traits elicitation, accurate and robust location for the trait point is convoluted and becoming a challenging issue in age prediction. Active Shape Model (ASM) can elicit the facial shape effectively and correctly. Furthermore, as the improvement of ASM, Active Appearance Models (AAM) is proposed to elicit both shape and texture traits from facial images simultaneously. In this paper, the two models are tested and compared for their performance against 6 algorithms which are Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Support Vector Regression (SVR), Canonical Correlation Analysis (CCA), Linear Discriminant Analysis (LDA), and Projection Twin Support Vector Machine (PTSVM). The experiments show that ASM is faster and gains more precise result than the AAM.