Age estimation from facial images based on hierarchical feature selection

The human face conveys remarkable amount of perceptible information and traits such as the emotional state, ethnicity, age and gender. In this paper, we explore a vision-based approach for the estimation of age range for an individual via facial features. The local binary pattern operator is applied to derive a hybrid set of features including local and global characteristics from the face. A histogram of features is constructed based on the concatenation of locally produced histogram vectors from grid cells. Hierarchical feature selection is described for the classification process where age ranges are classified in a tree-based fashion. Feature selection is based on the proximity of data instances belonging to the same age range is applied to obtain the most discriminative traits at each level of the defined age range. Experimental results carried out on a publicly available dataset confirmed the potency for the proposed method to better estimate the age range for different face images.