Abstract:

Automatic age estimation based on facial images has been become an important orientation of the face recognition research. By applying sparseness constrains to 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. And the radial basis function neural networks was used to extract the aging information contained in most facial image. The experimental results demonstrated that the non-negative matrix factorization with sparseness constraints algorithm could achieved a better performance for age estimation task.

Key words: age estimation; non-negative matrix factorization; sparse representation; facial image