OLPP-based Gabor feature dimensionality reduction for facial expression recognition

In facial expression recognition, high dimensional feature processing is still a hot topic since the solution to this problem can considerably reduce the time consuming operation and computational memory. Many methods have been developed to reduce feature dimension and extract the fundamental information in the feature space by projecting the original data into some lower dimensional space. In this paper, a method based on orthogonal locality preserving projections (OLPP), keeps the intrinsic structure properties of original data without losing the discriminated capacity simultaneously, is proposed to reduce high dimensional Gabor features for facial expression recognition. To evaluate the operation effectiveness, support vector machines (SVMs) is employed to classify such features, which are constructed by orthogonal bases obtaining from OLPP. Experiments are conducted to evaluate the performance of OLPP-based Gabor feature dimensionality reduction by comparison with locality preserving projections (LPP), linear discriminant analysis (LDA), and principle component analysis (PCA). Results demonstrate that OLPP outperforms the other three methods.