Towards Potential of N-back Task as Protocol and EEGNet for the EEG-based Biometric

Abstract

Electroencephalogram (EEG) has emerged as a biometric trait potentially with more security benefits compared to its conventional competitors such as fingerprint, iris scan, voice recognition, and face detection. However, there is still a long way to go to make EEG biometrics practical in real-world environments. One of the challenges of the EEG-based biometric systems is time efficiency. The protocols that can evoke individualdependent EEG patterns are usually time consuming. The signal-to-noise ratio (SNR) of the EEG signal is also low, which means a large number of epochs/trials (i.e. long acquisition time) are required to achieve a high accuracy recognition system. In this study we propose an EEG-based biometric model that could achieve high identification accuracy with data instances as short as only 1.1s (single epoch instances). In our biometric model, we propose a new protocol called the N-back task which is based on human working memory. As the nature of working memory is very short, it would be possible to elicit individual-dependent EEG responses within a very short period of time. The single epoch classification was achieved applying a deep neural network called EEGNet. Using 1.1s data instances, the proposed model could identify a pool of 26 subjects with the mean accuracy of 0.95, where recognition rate for majority of subjects was ≥0.99. Different components of this identity recognition model, from the proposed protocol to the classification algorithm, can be a line of research for the future of EEG biometric.