Age estimation using specific Gabor filter on upper facial area

Abstract

There is a growing research interest in facial aging estimation where the whole facial area was used for age estimation. However, in this paper, we propose a different approach for age estimation whereby we perform feature extraction at the specific part which is called the Region of Interest (ROI) on the upper facial area by employing a specific orientation and scales of Gabor filter. The proposed multi-Support Vector Machine (SVM) was used as the classification, and tested on two databases which are the captured face images of Malaysian citizen, and the FG-NET database. For the scheme, the Leave One Picture Out (LOPO) is used for training and testing according to age groups. The overall results of the proposed method show that the upper ROI performances are better for both Malaysian citizen and FG-NET database than the full facial ROI. Moreover, the Mean Absolute Error (MAE) for the FG-NET decreases when using the upper ROI approach. © 2013 Springer.