Recognition of Expressions Based on Kernel Global and Local Symmetrical Weighted Fisher Discriminant Nonlinear Subspace Approach

摘要

Preservation of global and local features of images during dimensional reduction is a challenging task. The main goal of this work is to resolve the problem of singularity matrix by preserving local and global discriminative features by introducing symmetrical weights on principal components. To meet this goal Combinational Entire Gabor Kernel Global and Locality Preserving Symmetrical Weighted Fisher Discriminant Analysis (CEGKGLSWFDA) approach is proposed. In this work extracted geometrical features are fused with Gabor magnitude and phase parts separately and these feature vectors are isolated. These isolated feature vector spaces are projected into subspace by KGLSWFDA method by enhancing and preserving the kernel discriminant global features in local domain. Matching score level post fusion technique is implemented on similarity score matrices of projected subspace. Based on final scores obtained from projected subspace of train dataset and test images expressions are recognized using Euclidean distance metric. Support vector machine (SVM) classifier is implemented to classify the expressions. Performance analysis is carried out by comparing earlier state of art approaches. Experimental results on JAFFE and YALE databases demonstrate the effectiveness of the proposed approach.