Biometrics such as fingerprints and iris scans has been used in authentication. However, conventional biometrics is vulnerable to identity theft, especially in user-management systems. As a new biometrics without this vulnerability, brain waves have been a focus. In this paper, brain waves (electroencephalograms (EEGs)) were measured from ten experiment subjects. Individual features were extracted from the log power spectra of the EEGs using principal component analysis, and verification was achieved using a support vector machine. It was found that, for the proposed authentication method, the equal error rate (EER) for a single electrode was about 22–32%, and that, for a multiple electrodes, was 4.4% by using the majority decision rule. Furthermore, nonlinear features based on chaos analysis were introduced for feature extraction and then extended to multidimensional ones. By fusing the results of all electrodes when using the proposed multidimensional nonlinear features and the spectral feature, an EER of 0% was achieved. As a result, it was confirmed that individuals can be authenticated using induced brain waves when they are subjected to ultrasounds.
This paper proposes the use of brain waves (EEGs) evoked by invisible visual stimulation as biometrics. To create invisible visual stimulations, we inserted a target image (a frame) into a video displayed at a high frame rate and simultaneously decreased the target intensity. The brain waves of experimental subjects presented with the invisible visual stimulations were measured. By comparing the measured brain waves with the knowledge obtained in conventional studies, we obtain the spectral differences under different intensity conditions. Using these spectral differences as individual features and the Euclidian distance matching as a verification method, we evaluated the verification performance of brain waves evoked by invisible visual stimulation. The equal error rate was minimized when using the spectra in the high-beta band.
This paper examines how to identify individuals using their electroencephalograms (EEGs) evoked by imperceptible vibration stimulation. Based on the knowledge of evoked EEGs by tactile stimulation, we learned that evoked responses could occur within a shorter time. In this paper, we propose measuring evoked EEGs for 100 ms. We confirm the presence of an event-related response in the measured EEGs and evaluate the individuality and verification performance of the measured EEGs as biometrics. The results demonstrate that the proposed stimulus presentation method is superior to the conventional method.