Feature-aging for age-invariant face recognition

Age-invariant face recognition has attracted some recent attention. In real applications, the age progression of those face images, stored in a face database for recognition and identification purposes, should also be considered, so as to achieve a higher accuracy level. In this paper, we propose a method to predict the aging of facial features so as to alleviate the effect of age progression on face recognition. The original facial feature and the aged facial feature of a face image should be correlated, so they are fused by using canonical correlation analysis to form a coherent feature for face recognition. The performance of our proposed approach is evaluated based on the FGNet database, and compared to some existing face recognition algorithms. Experiment results show that our proposed method can achieve a superior performance, when the query and probe face images have a large age difference.