Aging face identification using biologically inspired features

Aging face identification is a challenging task because facial aging process could degrade recognition performance dramatically. In this paper, we propose a method based on HMAX model and PCA to improve aging face identification performance. Since aging face images invariably differ in rotation, lighting and size, we normalize each face images to a standard size for minimizing these variations. After applying HMAX model and PCA on each normalized face image to get a new set of biologically inspired features (C1), we use it for aging face identification with the nearest neighbor rule and Mahalanobis distance from rank 1 to 6. Experimental results on six groups comprising of 10, 20, 30, 40, 50 and 52 pairs of face images from FG-NET database show that our method using C1 features could improve the identification accuracy around 3% instead of using raw face images.