Age Estimation Robust to Expression Variations

In this paper, we present a novel age estimation method that is able to deal with varying face expressions. We construct two efficient descriptors for face appearance which are robust to expression variations. One is termed as Proportion Descriptor constructed by proportion indices based on facial geometric features, the other is termed as Local Descriptor based on local facial texture features extracted by Modular Principle Component Analysis (MPCA). We set up experiments for evaluating the performance of the two facial descriptors in the presence of facial expressions.