Biometric Authentication using Evoked EEG by Invisible Visual Stimulation - Feature Extraction Based on Wavelet Transform -

In this study, we aim at the realization of authentication using evoked electroencephalogram (EEG) when presenting invisible visual stimulation as biometrics authentication towards safer and continuous authentication. In the previous researches, the measured EEG signal was processed by fast Fourier transform (FFT), and the power spectrum obtained was used as an individual feature, but the equal error rate (EER) representing the verification rate was about 43%. Therefore, in this paper, we introduce wavelet transform, which is a time-frequency analysis method, and extract a new individual feature including temporal information to improve the verification rate. As a result of evaluating the verification performance, in the case of presenting an invisible visual stimulation, the verification rate averaged over all electrodes tends to be improved as temporal information is included. In addition, as a result of evaluating the verification performance with data in which the start time of presenting stimulation is synchronized, the EER is the best at 14.0%, which is greatly improved compared to the conventional verification rate.