Ensemble-based machine learning models for vehicle drivers’ fatigue state detection utilizing EEG signals

Currently, there is a great extent of academic research focused on evaluating fatigue among drivers due to its growing recognition as a major contributor to vehicle tragedies. Combining advanced features and machine learning techniques, signals from the electroencephalogram (EEG) can be analyzed to...

Full description

Bibliographic Details
Main Authors: Hasan, Md Mahmudul, Islam, Md Nahidul, Khandaker, Sayma, Norizam, Sulaiman, Islam, Ashraful, Hossain, Mirza Mahfuj
Format: Article
Language:English
Published: University of Nis 2024
Subjects:
Online Access:http://umpir.ump.edu.my/id/eprint/43899/
http://umpir.ump.edu.my/id/eprint/43899/1/Ensemble-based%20machine%20learning%20models.pdf
Description
Summary:Currently, there is a great extent of academic research focused on evaluating fatigue among drivers due to its growing recognition as a major contributor to vehicle tragedies. Combining advanced features and machine learning techniques, signals from the electroencephalogram (EEG) can be analyzed to efficiently detect fatigue in the shortest possible time. This study presents an innovative approach to detect driver fatigue states utilizing ensemble-based machine learning techniques from EEG signals. Two ensemble models (Ensemble-based RUSBoosted Decision Trees and Ensemblebased Random Subspace Discriminant) were applied and compared. The study utilized an online EEG dataset of 12 individuals, with data collected during normal and fatigued driving conditions and Fast Fourier Transform was applied for feature extraction. The Ensemble-based RUSBoosted Decision Trees model achieved superior performance with 98.53% classification accuracy, compared to 83.13% for the Random Subspace Discriminant model. Multiple performance metrics were used for evaluation model performance. Finally, the proposed Ensemble-based RUSBoosted Decision Trees model outperformed Ensemble-based Random Subspace Discriminant model and existing conventional methods for fatigue state detection. This research contributes to the development of more accurate and reliable fatigue detection systems, which could potentially improve road safety by identifying fatigued drivers in real-time.