Styanidis A., Lourenço A., Carreiras C., Ahlström C., Ziakopoulos A., Yannis G., “Temporal Modeling of Heart Rate Variability for Driver Drowsiness Detection with LSTM Networks” at the Proceedings of the 8th IRTAD International Conference, Athens, 15-17 April 2026.

Drowsiness detection plays a crucial role in enhancing safety and preventing accidents in domains such as transportation and industrial operations. This work presents a data-driven approach for detecting driver drowsiness using electrocardiogram (ECG) signals from 20 sleep-deprived participants driving on a motorway in real-world driving conditions. Heart rate variability (HRV) features were extracted from ECG signals and used as sequential input to Long Short-Term Memory (LSTM) networks, which are well-suited for modeling temporal physiological patterns associated with drowsiness. The model was trained and evaluated using 10-fold cross-validation and leave-one-subject-out (LOSO) cross-validation strategies to ensure generalization across individuals. Experimental results demonstrate the model’s ability to distinguish between alert and drowsy states (10-fold accuracy = 87.8%, LOSO accuracy = 61.0%), highlighting the potential of HRV-based deep learning approaches for robust, non-intrusive drowsiness detection. This work underscores the importance of physiological signal analysis and deep learning techniques in developing reliable real-time driver monitoring systems.