Facial Recognition Using Square Diagonal Matrix Based on Two-Dimensional Linear Discriminant Analysis

Abstract

In this research, the square diagonal matrix based on Two-Dimensional Linear Discriminant Analysis for Face recognition is proposed. The original image matrix is converted into the diagonal matrix and followed the same process by using the output become input. The results of the conversion are utilized as input on feature extraction using Two-Dimensional Linear Discriminant Analysis. The proposed method has been evaluated by using the YALE-A, the ORL and the University of Bern Face image databases. The experimental results show that, the proposed method outperformed to other methods, which are Eigen Faces, Fisher Faces, Laplacian Faces and Orthogonal Laplacian Faces methods. The highest recognition rates are 89.523%, 97% and 95.33% for the YALE-A, the ORL, and the University of Bern respectively. The usage of the dominant features has influenced the results of the recognition. It is proved that, the more the dominant features, the smaller error occurred.