This study presents a face recognition system developed to address the problem of partial occlusions using a structured and targeted methodology. The proposed approach combines Principal Component Analysis (PCA) and Local Binary Patterns (LBP) for hybrid feature extraction, followed by a Support Vector Machine (SVM) classifier with an RBF kernel. The research was initiated through the observation that recognition models often fail when facial regions are obstructed. Through careful dataset design, parameter tuning, and confidence interval analysis, the final model demonstrates high recognition accuracy, minimal misclassification, and resilience to occlusions such as sunglasses, hair, hands on the face, and lighting variation. The approach achieves significant improvement over referenced benchmarks and contributes a statistically supported model that is lightweight and deployable in constrained environments.