Manifold learning method can discover intrinsic low-dimensional submanifold embedded in the high-dimensional image space

it attracts more and more attention in the research area of biometrics and cognitive science. However

manifold learning is an unsupervised learning method

the discriminative ability of the low-dimensional feature obtained by the algorithm is often lower than those obtained by the conventional dimensionality reduction methods. Furthermore

manifold learning methods don’t have direct mapping for new example

so it’s difficult to acquire the low dimensional features of new example. To address the two problems

this paper introduces a novel supervised kernel local linear embedding (SKLLE) method for facial expression recognition

it maps face images to high dimensional kernel space through nonlinear kernel mapping

then fuses prior class-label information and nonlinear facial expression submanifold of real face images to extract discriminative features for expression classification. SKLLE can not only gains a perfect approximation of facial expression manifold

but also enhances local within-class relations. It also does well on the new samples. Experimental results show that the proposed method can improve face expression classification performance effectively.