SUPERVISED ORTHOGONAL TENSOR NEIGHBORHOOD PRESERVING EMBEDDING FOR FACE RECOGNITION

Jianjun Chen

Yuanpei College of Shaoxing University, China

The deficiency of supervised discriminant information is the problem of Orthogonal Tensor Neighborhood Preserving Embedding (OTNPE) proposed recently for face recognition. So a dimension reduction algorithm called Supervised Orthogonal Tensor Neighborhood Preserving Embedding (SOTNPE) is proposed in the paper. On the basic of OTNPE, the algorithm achieves neighborhood reconstructions within the same class, preserving supervised class label information and neighborhood reconstruction information. Experiments on AR and YaleB face datasets show our proposed algorithm is efficient.