Facial Age Estimation Using Geometric, Local Phase Quantization and Pyramid Histogram of Oriented Gradient Features

Variations in shape and texture of a face are caused by aging across years. Factors such as genetics, lifestyle, health etc. greatly influence the aging process of a person which makes human age estimation from a facial image a very difficult problem. In this research, geometric and texture based facial feature extraction methods are presented. These methods are based on fusion between geometric features and local phase quantization (LPQ) texture features or geometric and pyramid histogram of oriented gradients (PHOG) texture features. Classification based on these methods was then carried out using the K-Nearest Neighbor (KNN) and support vector machine (SVM) to determine the accuracy of the system. Result obtained showed that fusing geometric and PHOG features provided best classification accuracy of 69.4% using the FGNET Database 1002 images of various age ranges.