In the realm of facial recognition-based attendance systems, the quest for enhanced accuracy encounters challenges. Existing methodologies grapple with accuracy issues, necessitating the exploration of robust techniques to address these limitations effectively. This study focuses on implementing a continuous video attendance system, departing from the conventional approach that captures attendance at fixed intervals, potentially missing faces. The proposed work produces attendance by using a histogram-oriented gradient (HOG) and local binary pattern histogram (LBPH) algorithms over specific time intervals, ensuring comprehensive and accurate daily attendance records. Incorporating this robust technique into facial recognition-based attendance systems not only overcomes existing limitations but also establishes a foundation for heightened precision and reliability in attendance management. The proposed method utilizes HOG, which outperforms well-known techniques such as the NN classifier, multikeypoint descriptors (MKD), the Gabor ternary pattern (GTP), noise-resistant LBP, principal component analysis (PCA), and deep learning. With an accuracy of 95%, the system shows potential for high-speed response and enhanced accuracy in real-world scenarios, especially with larger datasets.