EEG-Based Biometrics for User Identification Using Deep Learning Method

Humaira Nisar, Jia Yet Cheong, Vooi Voon Yap

Research output: Chapter in Book/Report/Conference proceedingConference Proceeding (Non-Journal item)

Abstract

Electroencephalogram (EEG) based biometric system is a relatively secure system that could be integrated into our daily lives replacing the existing biometric security systems that use fingerprints or facial features. In this paper, EEGNet has been used as a deep learning tool to build a biometric system by training the models using resting state raw EEG data. The dataset consists of 25 participants and the data was collected for 20 sessions over 10 weeks. The user identification accuracy obtained is 0.903 ± 0.090 and 0.923 ± 0.071 using 300 and 600 seconds of training data respectively. The effect of template aging had also been studied and found that the performance using 600 seconds for training dropped from 0.923 to 0.899 after one session and down to 0.847 after 14 sessions, which is around a 7 -week interval between training and testing of models.

Original languageEnglish
Title of host publication2024 IEEE 8th International Conference on Signal and Image Processing Applications (ICSIPA)
PublisherIEEE Press
Number of pages6
ISBN (Electronic)9798350352368
ISBN (Print)9798350352375
DOIs
Publication statusPublished - 03 Sept 2024
Event8th IEEE International Conference on Signal and Image Processing Applications, ICSIPA 2024 - Kuala Lumpur, Malaysia
Duration: 03 Sept 202405 Sept 2024

Publication series

NameIEEE International Conference on Signal and Image Processing Applications
PublisherIEEE Press

Conference

Conference8th IEEE International Conference on Signal and Image Processing Applications, ICSIPA 2024
Country/TerritoryMalaysia
CityKuala Lumpur
Period03 Sept 202405 Sept 2024

Keywords

  • Biometrics
  • Deep Learning
  • EEGNet
  • Raw EEG

Fingerprint

Dive into the research topics of 'EEG-Based Biometrics for User Identification Using Deep Learning Method'. Together they form a unique fingerprint.

Cite this