One of the major causes of road accidents is drowsy driving. Identifying drivers' drowsiness state is a crucial step in preventing accidents. The state of drowsiness of a driver can be assessed by analyzing the brain's electrical activity using electroencephalogram (EEG) signals. In this paper., the performance of pre-trained neural networks for automatic drowsiness detection using single-channel EEG spectrogram on PhysioNet sleep-EDF dataset is investigated. In addition., we propose a 1D-CNN to take advantage of the time-domain features of the EEG signals to improve the accuracy of the predictions. The performance and robustness of the proposed model are further improved using averaging and stacking fusion methods. Results of this study show that the proposed stacking-average fusion method provides the best level of accuracy (90.73%) for cross-subject test data. The proposed model simultaneously utilizes both the time-domain and frequency-domain features of EEG data., improving the automatic assessment of drowsiness.
Automated Driver Drowsiness Detection from Single-Channel EEG Signals Using Convolutional Neural Networks and Transfer Learning
2022-10-08
874121 byte
Conference paper
Electronic Resource
English
Driver Drowsiness Detection Using Deep Learning
IEEE | 2023
|Driver Drowsiness Detection Using Deep Learning
IEEE | 2021
|Driver Drowsiness Detection Using Machine Learning
IEEE | 2023
|