A Facial Authentication Method Robust to Postural Changes in e-Testing
説明
Examinee posture changes during e-Testing, making it difficult to perform facial authentication throughout the duration of the exam. We propose two methods for addressing this. The first method is to combine facial detection and tracking of facial information during the test. The second method is to pre-register facial information at various postures for authentication by computing the most similar image at the time of authentication. We conducted certification experiments using 21 examinees. The results showed authentication accuracy of 82.8%, an improvement of 16.8% over previous research, thus demonstrating the effectiveness of the proposed method.
