基于稀疏性约束非负矩阵分解的人脸年龄估计方法

Age estimation of facial images based on non-negative matrix factorization with sparseness constraints

基于人脸图像的年龄自动估计已经成为当前人脸识别领域的一个重要研究方向。首先通过非负矩阵分解(non-negative matrix factorization,NMF)算法对基矩阵或系数矩阵进行稀疏性约束,用形成的更具有局部表达能力的子空间对人脸图像数据进行表示。然后使用径向基函数神经网络进行训练和测试,提取包含在大多数人脸图像上的年龄信息来进行年龄估计。实验结果表明,具有稀疏性约束的非负矩阵分解算法对年龄估计问题具有良好的应用效果。

Automatic age estimation based on facial images has been become an important orientation of face recognition research.By applying sparseness constrains to the base matrix or coefficient matrix in the factorization of non-negative matrix factorization,a new subspace could be formed with part-based representation ability to describe image data.Radial basis function neural networks were used to extract the aging information contained in most facial images.The experimental results demonstrated that the non-negative matrix factorization with a sparseness constraints algorithm could achieve better performance for the age estimation task.