Biometric identification of persons using sample entropy features of EEG during rest state

Biometric recognition of persons using brain waves has been identified as an attractive topic of research today. Existing popular biometric modalities of face, finger prints and voice signals are vulnerable to various kinds of attacks and spoofing techniques, whereas the emerging biometric trait extracted from brain wave is expected to act as an ideal biometric feature offering high degree of uniqueness, stability and universality. This paper analyses the efficacy of the complexity of Electroencephalogram (EEG) signals recorded during rest state for recognizing individuals from a publicly available EEG dataset consisting of 109 subjects. Sample entropy features extracted from delta, theta, alpha, beta and gamma bands of 64 channel EEG have been evaluated for subject-identification in the proposed system. It is found that beta band entropy has the highest inter-subject variability. Based on a Mahalanobis distance based classifier, beta entropy gives an average correct recognition rate of 98.31%. It has also been observed that concatenation of entropy features with power spectral density (PSD) values improves the system performance. Further analysis is essential to investigate the stability of results over time and to optimize the recognition performance at a reduced number of channels.