Distance-weighted manifold learning in facial expression recognition
Du Jing, Li Bo · 2016
Facial expression recognition is a challenge problem in biometric identification. Many traditional methods, i.e. Local Preserving Projection (LPP) algorithm, have attempted to solve the problem. In this paper, a distance weighted supervised manifold learning algorithm based on LPP is also proposed. In the proposed method, a distance-weighted matrix, which fully takes advantage of the class information and degree of classes' diversity, is introduced. And then, the weighted LPP based on the distance-weighted matrix is applied to facial expression data. Compared with LPP, Unsupervised Discriminant Projection (UDP), Kernel Linear Discriminant Analysis (KLDA), experiment results on CK and JAFFE face database show that our method is effective and efficiency.