Face recognition system based on novel Peak Response-based Sobel Directional Pattern (PRSDP) and the Stacked Restricted Boltzmann Machine (SRBM)

Abstract-Face recognition systems are now extensively used in a variety of daily needs. Despite significant advances, occlusions, changes in light, camera types, and resolutions continue to pose challenges to face recognition systems. An effective face recognition system based on a Peak Response-based Sobel Directional Pattern (PRSDP) and the Stacked Restricted Boltzmann Machine (SRBM) technique was developed in this study. The major goal of the proposed system is the improved feature extraction approach to improve the features recovered by the PRSDP by removing noise. The Principal Component Analysis and Linear Discriminant Analysis (PCA + LDA) technique is then used to decrease the enhanced features. Experiments on JAFFE, LFW and Oulu-CASIA reveal that the suggested approach, when combined with Gabor and SRBM, can generate promising results and surpass previously proposed methods.