Face Recognition Based on Random Weights Network and Quasi Singular Value Decomposition
This paper proposes a novel approach of feature extraction called quasi singular values decomposition (QSVD), which can be used to obtain the algebraic features of the original images. An effective classifier, named random weights network (RWN), is applied to improve the learning speed. Integrating QSVD with RWN, fast discrete curvelet transform (FDCT), and 2-dimensional principal component analysis (2DPCA), a new method for face recognition is designed. The experimental results illustrate that the proposed method has an outstanding superiority in the aspects of separability and recognition rate.
KeywordsFace recognition Quasi singular value decomposition Random weights network Fast discrete curvelet transform 2DPCA
Unable to display preview. Download preview PDF.
- 2.Donoho, D.L., Duncan, M.R.: Digital Curvelet Transform: Strategy, Implementation and Experiments. In: Proceedings of SPIE, vol. 4056, pp. 12–30 (2000)Google Scholar
- 9.Wang, Y.H., Tan, T.N., Zhu, Y.: Face Identification Based on Singular Values Decomposition and Data Fusion. Chinese J. Comput. 23(6), 23–26 (2000)Google Scholar