一种基于人脸图像的年龄估计方法

有关年龄估计的研究在人机交互领域有着非常重要的意义。该文提出一种基于人脸图像的年龄估计方法,该方法首先基于颅面成长模式理论建立人脸测量模板,在此模板上计算面部几何比例特征,然后运用分数阶微分提取人脸局部区域的纹理特征,结合这两类特征构成个体年龄特征向量;通过聚类学习的方法训练年龄特征向量获得年龄-特征映射矩阵,最后由此矩阵表决出输入人脸的估计年龄。实验结果表明,基于这两种特征构建的年龄估计模型可以获得较好的年龄估计结果,年龄误差较小,分类准确率接近人的主观判断结果。

An Age Estimation Method Based on Facial Images

Research on age estimation has a significant impact on Human-Computer Interaction.In this paper,an age estimation method based on facial images is proposed.The new method establishes a face anthropometry template based on craniofacial growth pattern theory to obtain facial geometric proportion features,and extracts texture features of facial local area using fractional differential approach,combines these two kinds of features to form personal age feature vectors.Machine learning methods such as clustering algorithms we used to obtain age-feature knowledge matrix,and in age estimating,such knowledge matrix voting on estimate age of input facial image.Experi-mental results show that the estimation error is small and the classification accuracy is close to human judgment.