CNN-LSTM To Identify The Most Informative EEG-Based Driver Drowsiness Detection Brain Region

• Published in: 2022 International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT) pp. 725 - 730
• Main Authors: Latreche, Imene, Slatnia, Sihem, Kazar, Okba
• Format: Conference Proceeding
• Language: English
• Published: IEEE 20-10-2022
• Subjects: Brain modeling; Brain regions; CNN; Costs; Deep learning; Driver Drowsiness; Electrodes; Electroencephalogram (EEG); LSTM; Scalp; Signal processing; Wearable computers
• ISSN: 2770-7962
• DOI: 10.1109/ISMSIT56059.2022.9932696

As driver drowsiness causes dangerous results such as accidents and deaths, it is crucial to develop a system that can detect it in its early stage and precisely. Numerous studies employed various methods to detect drowsiness; however, Electroencephalogram (EEG) signals were the most accurate. A driver drowsiness detection system will be placed in a vehicle to detect it in real-time, so a wearable device is required to acquire EEG signals from the driver's scalp. However, a wearable device with a large number of electrodes may be uncomfortable for the driver. In addition, it is expensive in terms of cost and complexity of analysis. Our study aims to identify the most informative brain region to overcome this issue without losing essential information. We have used a deep learning model with an online available dataset to define the accurate region. The results are validated with a ten-fold cross-validation technique. The experiment results indicate that the best single-region is the central region with a mean accuracy of 73.55% and three electrodes, and the best region combination was the central-occipital (CO) with an accuracy of 75.53% and six electrodes which is a prominent result that can enhance the EEG-based drowsiness detection rate.