Impact of fatigue levels on EEG-based personal recognition

Area of Science:

• Neuroscience
• Biometrics
• Human-Computer Interaction

Background:

• Electroencephalogram (EEG) offers unique biometric markers for user authentication in brain-computer interface (BCI) systems.
• User state fluctuations, particularly fatigue, can negatively impact EEG-based biometric recognition performance.
• The impact of fatigue on individual recognition accuracy in EEG systems requires further investigation.

Purpose of the Study:

• To investigate the effects of varying fatigue levels on the performance of EEG-based personal recognition systems.
• To analyze how fatigue influences identity recognition accuracy within and across different fatigue states.
• To identify robust features for EEG-based biometrics that are resilient to fatigue.

Main Methods:

• Derived six sub-datasets from simulated driving data, each labeled with distinct fatigue levels.
• Extracted six features from each sub-dataset for identity recognition analysis.
• Conducted single-session and cross-session studies to evaluate recognition accuracy under varying fatigue conditions.

Main Results:

• Recognition accuracy declined significantly, dropping by over 30% after 90 minutes of simulated driving.
• System performance degraded more when testing on less fatigued data after training on fatigued data.
• Functional connectivity features demonstrated superior recognition accuracy across different fatigue levels.

Conclusions:

• Fatigue is a critical factor affecting EEG-based biometric recognition, necessitating its consideration in system design.
• Training datasets incorporating fatigued EEG states can improve cross-session recognition performance.
• Functional connectivity analysis presents a promising approach for developing fatigue-resilient EEG biometric systems.