Exploring Permanence and Uniqueness of EEG Brain Signals as a Biometric Signature, Part-II: Statistical Techniques

In this second part of our study, we continue our exploration of EEG-based biometric authentication by employing both parametric (T-test) and non-parametric (Mann-Whitney U-test) statistical techniques to measure the similarity of time-series signals. After filtering stimuli response signals, we employed two distinct normalization techniques-Forward-Backward-Slope-Normalization and Median-Referenced-Normalization. Additionally, two time-domain filters- Adjusted Laplacian of Mean and Laplacian of Gaussian - are applied for extraction of specific spectral neural activity. Through this multifaceted approach, we aim to enhance our understanding of the robustness and effectiveness of EEG-derived biometric signatures, offering valuable insights for future applications in this promising field. Our preliminary findings indicate that the proposed statistical-based pipeline can achieve an 86% accuracy rate in identifying individuals over the course of a week. This suggests that the pipeline shows promise in its ability to reliably distinguish between individuals based on the data it processes. However, further validation and testing may be necessary to confirm and refine these findings.