Hybrid LBPH-SVM For Face Recognition of Class Attendance

The conventional method of manually recording attendance in educational settings presents intrinsic difficulties, thereby demanding a more streamlined and precise alternative. Nevertheless, the implementation of a robust facial recognition system in an unconstrained setting, fraught with variables such as fluctuating facial orientations and expressions, unpredictable and suboptimal lighting circumstances, as well as diminished image clarity, continues to pose a formidable challenge. This project presents an innovative solution to tackle the issue at hand by introducing an automated system powered by advanced facial recognition technology. The use of the Hybrid LBPH-SVM model, which permits accurate face detection and recognition, is the main topic of the study. First, the LBPH identifies students’ faces from different angles to obtain the face image features, and finally employs SVM, to execute the classification. This proposed model can identify student regardless of their orientation towards the camera, thus significantly enhancing the system's efficiency and effectiveness. The experiments utilized the UiTM student dataset as a case study and results indicate an astounding 95% accuracy rate in lighting condition tests, which outperforms other approaches in distance testing, which have an average accuracy of 90%. By verifying the LBP face identification model's efficacy, this study significantly contributes to the existing pool of knowledge.