EEG-based biometric identification with deep learning

Abstract

Despite the recent increasing interest in biometric identification using electroencephalogram (EEG) signals, the state of the art still lacks a simple and robust model that is useful in real applications. This work proposes a new approach based on convolutional neural network CNN. The proposed CNN works directly on raw EEG data, thus alleviating the need for engineering features. We investigate the performance of the CNN on datasets of 100 subjects collected from one driving fatigue experiment. Our results show that the CNN model is fast highly efficient in training ( 100K training epochs) and highly robust, achieving 97% accuracy in identifying ∼14K testing epochs from 100 subjects with non-time-locked natural driving fatigue data and much higher than from randomly sampled epochs (90%). Overall, this work demonstrates the potential of deep learning solutions for real-life EEG-based biometric identification.