EEG biometrics based on small intra-individual and large inter-individual difference of extracted features

Biometrics refers to the process of identifying an individual from others by biological means. Most of the biometric systems are unreliable, can be imitated or even can be stolen. As a result, we need to search for a new biometrics and Electroencephalogram (EEG) based biometrics is a promising field in this aspect. By using the small intra-individual and large inter-individual difference in features with different trials, individuals can be identified with more accuracy. In this paper, a methodology for identifying an individual is proposed by determining the most effective domain and features of EEG signal. Three feed forward, back propagation multi-layer neural networks were built using the most effective features. The relative comparison shows that the network designed using the features of time domain gives the worst performance whereas the network designed using the features of both time and frequency domain gives the best performance for identifying an individual having relatively lower mean square error.