Dual-space neighborhood discriminant embedding for face recognition

作者：Liu Zhonghua^*; Zhou Yuan; Jin Zhong

来源：Journal of Computational Information Systems, 2011, 7(4): 1100-1107.

• Discriminant embedding
• Face data
• Face database
• Face images
• Face space
• Facial Expressions
• Gabor feature
• Gabor wavelet representation
• High-dimensional
• Local structure
• Locally linear embedding
• Low-dimensional subspace
• Neighborhood discriminant embedding
• Novel algorithm
• Pattern classification
• Small sample size problems
• Subspace learning
• Training sample

In this paper, a novel subspace learning method called neighborhood discriminant embedding (NDE) is proposed for pattern classification. In our algorithm, the neighbor and class relations of training samples data are used to construct the low-dimensional embedding submanifold. After being embedding into a low-dimensional subspace, in a local structure, samples from the same class will be as close as possible, and those from different classes are separated far way. However, NDE will suffer from the small sample size problem when dealing with the high dimensional face data. In order to solve the small sample size problem and take fully advantage of the discriminative information in the face space, we further propose a novel algorithm called dual-space neighborhood discriminant embedding (DSNDE). Since Gabor wavelet representation of face images is robust to variations due to illumination and facial expression changes, we apply the proposed DSNDE (NDE) algorithm on Gabor features for face recognition. Experiments on the face databases demonstrate the effectiveness of our method.

• 出版日期2011
• 单位河南大学