FACE RECOGNITION ACROSS AGES

This paper is concerned with the effect of ageing on biometric systems and particularly its impact on face recognition systems. Being biological tissue in nature, facial biometric trait undergoes significant changes as a person ages. Consequently, developing biometric applications for long-term use becomes a particularly challenging task. The idea behind the investigation presented here is that biometric systems have uneven difficulty in recognising people from different ages. Some algorithms may perform better for certain age groups. Therefore, a carefully optimised multi-algorithmic system can reduce the error rates. A subset of 100 subjects from the MORPH-II database has been selected to test the performance of a face verification system. The population is split into 5 age bands (?19, 20-29, 30-39, 40-49, ?50 years) based on their age during enrolment. The facial image database used in the experiments here contains images acquired over a period of five years. In the proposed multi-classifier scheme, features extracted from face images are transformed by different projection algorithms prior to matching. It has been observed that all the age groups showed improved performances when compared to the single classifier error rates. Of all the groups, the EER were highest for the younger population (?19 year olds).