Online Electroencephalogram (EEG) based biometric authentication using visual and audio stimuli

Biometric recognition of individuals has been widely employed in establishing secure and trustworthy systems nowadays. However, due to its increased demand and usability, the associated security risks have been increased, which necessitates finding of more robust biometric traits than existing modalities. Recently, brain signal recorded by Electroencephalogram (EEG) technique has been reported as a potential biometric candidate on account of its high degree of uniqueness, stability and universality. This paper presents an EEG-based biometric authentication system employing brain patterns in response to a number of visual or auditory stimuli by seeing/hearing self, familiar and unfamiliar faces/voices. The system employs power spectral density (PSD) features extracted from alpha, beta and gamma bands of EEG for biometric authentication. The PSD values of multi-band EEG signals from 14 channels form template feature vector for each subject, and are stored in the database during enrollment phase. During online authentication, test feature vector is correlated with the respective template vector and the obtained correlation value is compared with a pre-defined threshold value. Based on the authentication experiments performed on 5 healthy subjects, the proposed system offers an overall accuracy of 79.73% with a false acceptance rate (FAR) of 13.91% and false rejection rate (FRR) of 26.6%.