BRAINPRINT AUTHENTICATION UNDER VARYING ENVIRONMENTAL CONDITIONS: MACHINE LEARNING VERSUS DEEP LEARNING
This study explores the potential of using EEG signals for biometric authentication through the development of Convolutional Neural Network (CNN) model. In particular, the electroencephalogram (EEG) signals were recorded in different ambient noise settings: quiet environment, low-distraction envir...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
SCIK Publishing Corporation
2026
|
| Subjects: | |
| Online Access: | http://ir.unimas.my/id/eprint/51579/1/9617-21134-1-PB.pdf http://ir.unimas.my/id/eprint/51579/ https://scik.org/index.php/cmbn/article/view/9617 https://doi.org/10.28919/cmbn/9617 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study explores the potential of using EEG signals for biometric authentication through the development
of Convolutional Neural Network (CNN) model. In particular, the electroencephalogram (EEG) signals were recorded
in different ambient noise settings: quiet environment, low-distraction environment, and high-distraction environment.
Traditionally, EEG-based authentication requires a separate feature extraction step prior to the use of machine learning
algorithm. Feature extraction process is usually cumbersome, which relies heavily on human experts, and prone to
information loss. Thus, a CNN model based on EEGNet architecture is proposed to train EEG datasets collected from 45 volunteers who were instructed to look at images presented to them in all the three acoustic conditions. Using a variety of performance metrics, notably precision and recall, the model’s performance was compared across various classification thresholds to account for the imbalanced nature of the dataset. The performance was also compared across different environmental conditions, with the highest F1-score in quiet conditions. Additionally, the CNN’s performance was compared against a probability-based Incremental Fuzzy-Rough Nearest Neighbour (prob-IncFRNN)
model, with former outperforming the latter in all metrics. |
|---|