An enhancement of the Eigenface algorithm using weber local descriptor applied in attendance management system

This study presents an improved face recognition system tackling the Eigenface algorithm's limitations regarding lighting variance, class separability, and classification. The proposed method incorporates Weber Local Descriptor (WLD) for illumination normalization during training and recognition. Further improvements include Kernel Principal Component Analysis (KPCA) for non-linear feature transformation, Linear Discriminant Analysis (LDA) to maximize class separability, and Ridge classification for noise-resistant recognition, replacing Euclidean distance. Testing on the extended Yale B dataset showed a significant accuracy increase from 5.63% (original Eigenface) to 99.83% (enhanced Eigenface). Evaluation on a custom dataset simulating real-world conditions (varying light, expressions) yielded 100% accuracy across feature transformation, class separability, and classification. These results demonstrate the effectiveness of the integrated WLD, KPCA, LDA, and Ridge classification techniques in developing a robust and accurate face recognition system suitable for applications like attendance management.