03486ntm a2200337 i 4500 MY-KuUP 20251230175348.0 t||||fr|||| 000 0 ta 251230t20252025my a|||fr|||| 000 0 eng d THE0010194 (Local) Hardback UMPSA eng UMPSA rda FTKEE .M34 2025 r Thesis Md Mahmudul Hasan, author. Hybrid bigru-bilstm with multi-head attention for eeg-based microsleep detection / Md Mahmudul Hasan Kuantan, Pahang : UMPSA, 2025 © 2025 xv, 133 pages : illustrations ; 30 cm. + 1 CD-ROM. rdacontent text rdamedia unmediated rdacarrier volume rda text file PDF Faculty of Electrical & Electronics Engineering Technology Thesis (Master of Science) -- Universiti Malaysia Pahang – 2025 Include bibliographical reference Microsleep poses a serious risk to high-risk occupations like driving, flying, and industrial work. Microsleep can result in serious accidents and fatalities and are frequently brought on by exhaustion, irregular circadian rhythms, or tedious work. This emphasizes the critical need for precise and timely detection systems. The accuracy, versatility, and practicality of current detection techniques are still constrained, despite developments in related domains. By putting forth an intelligent EEG-based system for early microsleep identification and alarm production, this study overcomes these constraints. This study presents a thorough detection framework that uses the sensitivity of electroencephalogram (EEG) signals to identify unique cortical brain activity patterns. To guarantee robustness, the process combines both controlled experimental data and publicly available EEG datasets, starting with data gathering utilizing an eight-channel Unicorn Black EEG headset. Utilizing sophisticated feature extraction methods, significant insights are extracted from unprocessed EEG signals. To improve detection accuracy, both conventional machine learning models including Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) and deep learning models including Convolutional Neural Network (CNN), Artificial Neural Network (ANN), and a new Hybrid BiGRU-BiLSTM with Multi-Head Attention are used. Comparative evaluations demonstrate the higher performance of the suggested hybrid design, which attains 98.89% classification accuracy, 99.16% specificity, 97.77% MCC, 99.05% precision, 98.58% recall, and 98.81% F1-score. Further evidence of its generalizability comes from validation on a publicly available dataset. Additionally, the study offers a hardware feedback system prototype and an intuitive graphical user interface (GUI) for real-time driver alertness monitoring and accident prevention. In summary, this study fills important gaps in existing microsleep detection techniques and offers a scalable solution that can be implemented in safety-critical areas. Future approaches include investigating cross-domain applications in industrial safety systems, transportation, and aviation, as well as diversifying datasets and improving hardware integration. Faculty of Electrical & Electronics Engineering Technology Dissertations Universities and colleges Dissertations Theses Dissertations lcc THESIS 0 0 lcc 0 1 20000 20000 2025-12-30 0 FTKEE .M34 2025 r Thesis T000003891 2025-12-30 17:55:45 2025-12-30 THESIS 0 0 lcc 0 -2 20000 20000 2025-12-30 0 CD13816 T000003892 2025-12-30 17:56:57 2025-12-30 THESIS 103813 103819