(英) |
Personal authentication using the physical characteristics, e.g. face and fingerprint, is vulnerable to spoofing although accuracy is high. Personal authentication using behavioral characteristics, e.g. signature, gait, and eye movement is robust against spoofing although accuracy is low. In this paper, we evaluated the effect of the presence or absence of memorized image (key-image) on the accuracy of eye movement authentication. We employed Complex Eye Movement feature (CEM), Cepstrum-based feature (CEP), and Eye Movement Local Binary Pattern (EMLBP) in the experiment. From the experimental results, there was no difference of the accuracy for personal authentication with or without key-images. The accuracy was improved by score level fusion of each feature. |