基于集成的年龄估计方法

A New Age Estimation Method Based on Ensemble Learning

近十年来,由于广泛的应用前景,关于人脸识别的研究得到了广泛的关注.但目前有一种影响人脸识别技术的因素尚未被研究者所重视,那就是年龄变化.而在适用于年龄变化的人脸识别技术中有一个重要的问题,即年龄估计.本文基于典型相关分析和代价敏感学习提出了两种年龄估计算法,并在此基础上利用集成技术来提高年龄估计的准确性.最终实验结果验证了本文方法的有效性.

For the last decade, research on face recognition has attracted much attention due to its wide application. However, the factor that affects the performance of face recognition and has been neglected by researchers is the aging variation. Age estimation is an important problem in the face recognition technique. This paper proposes two algorithms for age estimation by utilizing canonical correlation analysis and cost-sensitive learning. Furthermore, these two algorithms are combined into the ensemble method to improve the accuracy of age estimation. Finally, the experimental results show the effectiveness of the proposed method.