摘要
人脸年龄自动估计在人机交互中有着非常广阔的应用前景,正吸引着人们的广泛关注。然而,人脸年龄估计仍是一个极具挑战性的问题,为此,提出了一种新的年龄估计方法。首先,采用局部定向模式(LDP)和Gabor小波变换分别提取人脸的全局和局部特征。然后,基于信息融合理论对这两种特征进行融合,并用PCA方法对融合后的特征进行降维,从而获得低维的年龄特征向量。最后,利用支持向量回归(SVR)方法进行年龄估计。在公共的FG-NET年龄数据库上进行了实验,实验结果表明,所提出的方法是有效的。
Automatic facial aging estimation, which has the potential for many applications in human-computer interactions, has been receiving wide attention. However, the age estimation problem is challenging. A new method for facial age estimation is presented. Firstly, the global and local facial features are extracted by taking the local direction pattern （LDP） and Gabor wavelet transform respectively. Then, these features are combined according to the information fusion theory, and the PCA method is applied to further reduce the aging features dimension, so as to obtain the age feature vectors with low dimension. Finally, using support vector regression （SVR） method for age estimation. Experiments have been conducted at the public FG-NET age databases, experimental results show that the method proposed is effective.
出处
《科学技术与工程》
北大核心
2014年第3期76-80,共5页
Science Technology and Engineering
基金
广东省产学研合作项目(2012B091100043)资助
关键词
年龄估计
局部定向模式
信息融合
支持向量回归
age estimation local alrectlon pattern lnlormatlon IUSlOn support vector regression
