DefinePK

Abstract

The paper will present the most known techniques for feature selection and size reduction of vectors. Some linear and nonlinear techniques are designed and implemented such as Kernel Principal Component Analysis, Locally Linear Embedding, MPPCA, Generalized Discriminant Analysis, Laplacian Eigen-maps, Isomap, Landmark Isomap, and LTSA approaches. These approaches are applied and tested on different common biometric database, such as DRIONS, VARIA, IIT Delhi Ear and STORE datasets. The experimental results of the suggested techniques are presented and compared.