Biometric person authentication using new spectral features of electroencephalogram (EEG)

The electroencephalogram (EEG) is suitable for continuous authentication since it enables unconscious presentation of biometric features. However, the conventional authentications methods using the EEG require heavy computational load: AR modeling and/or machine learning. In this paper, we propose simple spectral features of the EEG to make the authentication system using the EEG more practical. The effectiveness of the proposed new features is evaluated in verification experiments with 23 users and we obtain the verification rate of 79%.