Manifold Adaptive Kernel Nonnegative Matrix Factorization for Face Recognition

作者：Sun Xia^*; Wang Ziqiang; Sun Lijun

来源：Journal of Computational and Theoretical Nanoscience, 2015, 12(9): 2710-2719.

DOI：10.1166/jetn.205.4166

• Face Recognition
• Nonnegative Matrix Factorization
• Manifold Structure
• Kernel Learning

Nonnegative matrix factorization (NMF) has become a popular dimensionality reduction method for face recognition due to its parts-based representation and nonnegativity constraints. However, NMF is a linear technique and ignores both the local manifold structure and discriminative information of face images. To overcome the aforementioned problems and enhance the performance of NMF in face recognition, a novel manifold adaptive kernel nonnegative matrix factorization (MAKNMF) algorithm for face recognition is proposed in this paper, which can explicitly consider the intrinsic manifold structure in face image space and yield nonlinear discriminating feature of the face image. A convergency provable multiplicative update rule is then derived to obtain the nonnegative basis matrix and transformation matrix. Experimental results on three well-known face image databases demonstrate the proposed MAKNMF algorithm outperforms other related algorithms in terms of recognition accuracy.

• 出版日期2015-9
• 单位河南工业大学