Author(s)

Yiting Shen

Corresponding Author

Yiting Shen

ABSTRACT

In order to improve the efficiency and accuracy of kernel principal component analysis, two novel improved methods of KPCA are come up with in this paper. One is the weighted KPCA (WKPCA) algorithm, the other is a face recognition method DKPCA based on discrete cosine transform (DCT) and kernel principal component analysis (KPCA). The former weights the projection matrix to improve the recognition rate, meanwhile, the latter performs DCT transformation on the database, selects low-frequency coefficients to reconstruct the face, extracts eigenvalues by KPCA, and adopts the nearest neighbor method for classification. The research results demonstrate that the two improved methods proposed in this paper can indeed improve the work efficiency and recognition accuracy of kernel principal component analysis.

KEYWORDS

Face recognition, KPCA, WKPCA, Projection matrix, k-Nearest Neighbor