Deep Learning-Based Attention Mechanism for Automatic Drowsiness Detection Using EEG Signal

• Published in: IEEE sensors letters Vol. 8; no. 3; pp. 1 - 4
• Main Authors: Divvala, Chiranjevulu, Mishra, Madhusudhan
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.03.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Alertness; Attention; Brain modeling; Convolution; convolutional neural networks (CNNs); Data models; Deep learning; deep learning (DL); electroencephalograph (EEG); Electroencephalography; Fatigue; Feature extraction; health care; Lethargy; Sensor systems; Sleep; Sleepiness; wearable sensor data
• ISSN: 2475-1472; 2475-1472
• DOI: 10.1109/LSENS.2024.3363735

An electroencephalograph (EEG) is the basic medical tool to identify disorders related to brain activity. Drowsiness is a natural signal from the body indicating the need for rest and sleep to restore physical and mental well-being. Drowsiness is characterized by lethargy, fatigue, and a strong inclination toward sleep. It is often accompanied by reduced alertness and increased difficulty in maintaining attention and focus on tasks. Individuals experiencing drowsiness may find staying awake challenging and exhibit slower reaction times. This diminished cognitive function can lead to accidents, errors, and decreased performance in various activities. Wearable sensors are utilized in real-time to identify drowsiness detection. However, an automated diagnosis tool is very helpful in identifying drowsiness, and detection is an important task. Therefore, this work proposes a deep learning-based attention mechanism to detect the drowsiness state. This letter uses a publicly available MIT-BIH standard EEG database for experimentation. The proposed model provides a performance accuracy of 98.38% in drowsiness detection. The experiment outcomes demonstrate an enhanced detection capability when compared with current state-of-the-art methods for detecting drowsiness using single-channel EEG signals.