Summary
International Symposium on Nonlinear Theory and its Applications
2010
Session Number:B1L-B
Session:
Number:B1L-B4
Age Estimation using Kernel Regression Analysis
Hironobu Fukai, Hironori Takimoto, Yasue Mitsukura, Minoru Fukumi,
pp.221-224
Publication Date:2010/9/5
Online ISSN:2188-5079
Summary:
In this paper, we propose age estimation using kernel regression analysis. Age estimation is one of most difficult problem of facial recognition research areas. However, if age can be estimated by computer, it is possible to apply it to various fields, for example, marketing, communication person and machine, and so on. Then, there are many researches, but practical use research is few. Furthermore, a necessary technology changes greatly also under the environment respectively, for example, use ID photo, use images of takes before laptop PC, use images that takes surveillance camera. We focus on images from surveillance camera, and we aim to propose generalized age estimation method. Therefore, we propose age estimation method using wrinkle and pigmented spot information that can be extract age feature invariably even if appearance changes. To extract pigmented spot and wrinkle information, we use ε-filter. Moreover, the kernel regression analysis is used for age estimation method. To evaluate effectiveness of the proposed, we simulate age estimation by using actual face image. As a result, age estimation error value is about 6.65 years old. This result is high accuracy as a generalized method.