A survey : linear and nonlinear PCA based face recognition techniques

Abstract EN

face recognition is considered to be one of the most reliable biometric, when security issues are taken into concern.

For this, feature extraction becomes a critical problem.

Different methods are used for extraction of facial feature which are broadly classified into linear and nonlinear subspaces.

Among the linear methods are Linear Discriminant Analysis (LDA), Bayesian Methods (MAP and ML), Discriminative Common Vectors (DCV), Independent Component Analysis (ICA), Tensor faces (Multi-Linear Singular Value Decomposition (SVD)), Two Dimensional PCA (2D-PCA), Two Dimensional LDA (2DLDA) etc.

but Principal Component Analysis (PCA) is considered to be one the classic method in this field.

Based on this a brief comparison of PCA family is drawn, of which PCA, Kernel PCA (KPCA), Two Dimensional PCA (2DPCA) and Two Dimensional Kernel (2DKPCA) are of major concern.

Based on literature review recognition performance of PCA family is analyzed using the databases named YALE, YALE-B, ORL and CMU.

Concluding remarks about testing criteria set by different authors as listed in literature reveals that K series of PCA produced better results as compared to simple PCA and 2DPCA on the aforementioned datasets.