Real-Time EEG Signal Analysis for Microsleep Detection: Hyper-Opt-ANN as a Key Solution

Microsleeps are brief lapses in awareness that pose significant risks, particularly in activities requiring continuous attention, such as driving. These episodes are common in sleep-deprived individuals and can lead to catastrophic outcomes. Electroencephalography (EEG) is a promising technique for detecting microsleeps due to its high temporal resolution, allowing real-time brain activity monitoring. The study aims to develop a lightweight version of the model to reduce computational costs and provide faster detection, enabling quicker intervention to prevent accidents in safety-critical environments. We propose a customized deep learning model, Hyper-Opt-ANN, designed to detect microsleep episodes from EEG signals. The model is evaluated across five time windows (1 second, 2 seconds, 3 seconds, 4 seconds, and 5 seconds), with the 4 seconds window showing the best performance. The Hyper-Opt-ANN model achieved a significant accuracy of 97.33%, demonstrating its efficacy and potential for accurate microsleep detection using EEG signals. This method significantly outperforms traditional approaches and has potential applications in safety-critical domains. This study demonstrates the feasibility of using EEG signals and advanced deep learning models for detecting microsleep and enhancing safety in high-risk environments.