A novel method of recognizing ageing face based on EHMM

Abstract

The existing automatic methods of face recognition cannot recognize ageing faces with great changes in facial appearance. In this paper, a novel algorithm based on EHMM (embedded hidden Markov model) is presented to recognize the face with large ageing effects. Firstly, the non-linear relations between age and motions of key feature points in face are achieved by analyzing a great number of samples, and then a non-linear model is built to describe these relations. Secondly, together with PARI (partial ageing ratio image), texture of the ageing faces is predicted. Thirdly, features extracted from reconstructed images are used to train EHMM and face recognition is performed in the test set, where faces are in large age span. Experimental result shows that the algorithm proposed outperformed the traditional EHMM method.