Comparison of Face Recognition Using PCLDA and Neural Network

Facial recognition is a complex multidimensional structure that demands sophisticated computing techniques for authentication purpose. In this paper, we introduce the Integral Normalized Gradient Image (INGI) algorithm with various normalizing stages. The system comprises a novel illumination insensitive preprocessing method, a hybrid Fourier based feature extraction and matching process. The Pre-processing method is grounded in the analysis of the facial imaging model, considering intrinsic and extrinsic factors of the human face. Feature extraction encompasses hybrid Fourier features extracted from different frequency bands and multiple face models. By deriving Fourier features from three Fourier domains and three distinct frequency bandwidths, we acquired additional complementary features. These features are individually classified using Principal Component and Linear Discriminant Analysis (PCLDA). This approach enables in analyzing a face image from the various viewpoints for identity recognition. Furthermore, we propose multiple face models based on different eye positions with a same image size. This contributes significantly to enhancing the performance of the proposed system. Recognition is achieved through Euclidean Distance and Neural Network based classifier, resulting in a recognition accuracy of approximately 89.23% for the Euclidean Distance classifier-based model and 93.40% for Back Propagation Neural Network Classifier.