Authors: Arsath Mohamed M, Gokul M, Rajamanikandan R, Mr.M.Premkumar
Abstract: Manual attendance recording consumes classroom time and may permit proxy attendance. Face-recognition-based attendance systems can automate identity matching, but their reliability may be affected by illumination, image quality, and presentation attacks. This paper presents ATTENDAI, a webbased student attendance system designed to combine face recognition, voice verification, and eye-blink liveness detection in a staged verification workflow. The proposed implementation uses Python and Flask for application services, OpenCV with Haar Cascade face detection and Local Binary Patterns Histograms(LBPH) for face processing, Mel-Frequency Cepstral Coefficients (MFCC) for voice features, and facial landmarks with the Eye Aspect Ratio (EAR) for blink detection. Attendance is recordeonly after the configured verification stages succeed. The system also supports Admin, Staff, and Student roles, departmentwise management, attendance reporting, and a four-hour session rule intended to prevent duplicate entries while retaining historical records. This paper describes the motivation, system architecture, principal modules, session logic, and an evaluation plan. Quantitative performance results are not claimed because they must be obtained from controlled testing of the completed implementation.
DOI: https://doi.org/10.5281/zenodo.23337736
