Citation

Fahim, Abu Monsur Mohammad and Rifat, Md Faisal Hoque and Tushe, Ummay Ayman and Rabbi, Riadul Islam and Tusher, Ekramul Haque and Liew, Tze Hui (2025) A Novel Three-Tier Driver Drowsiness Detection Framework Using Stress-Proxy HRV Analysis. In: 8th 2025 International Conference on New Media Studies, CONMEDIA 2025, 14 October 2025 - 17 October 2025, Malacca, Malaysia.

Text IEEE Xplore Full-Text PDF_5.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

Driver drowsiness remains a critical safety concern, contributing significantly to traffic accidents worldwide. While traditional detection methods rely on intrusive sensors or computationally intensive video processing, this study explores a non-invasive approach using comprehensive heart rate variability (HRV) analysis for multi-level drowsiness classification. This research utilizes the SWELL dataset to develop a novel three-class drowsiness detection framework, categorizing driver states into Alert, Early Drowsiness, and Severe Drowsiness levels through strategic mapping of stress conditions as drowsiness proxies. The methodology employs 33 comprehensive HRV features encompassing time-domain measures (MEAN RR, RMSSD, pNN50), frequency-domain parameters (VLF, LF, HF components), and nonlinear complexity metrics (sample entropy, Higuchi fractal dimension). A feed-forward artificial neural network with optimized architecture consisting of 64 and 32 hidden neurons processes standardized feature vectors to perform multi-class classification using softmax activation. The proposed system demonstrates promising performance in distinguishing between different drowsiness severity levels, offering advantages in real-time applicability due to HRV’s continuous availability and minimal computational requirements. Unlike binary alert/drowsy classifications common in existing literature, this multi-level approach provides granular drowsiness assessment, enabling graduated warning systems. The non-intrusive nature of HRV monitoring through wearable devices makes this approach practically viable for real-world automotive integration. Key contributions include the novel application of stress-condition proxy mapping for drowsiness classification, comprehensive feature engineering combining multiple HRV domains, and validation of multi-level drowsiness detection feasibility. Future enhancements could incorporate driver-specific personalization to address interindividual HRV variations, integration with additional physiological signals, and embedded system implementation for real-time automotive deployment. This research advances the development of practical, non-invasive drowsiness monitoring systems with potential for significant road safety improvements.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Artificial neural networks
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Information Science and Technology (FIST)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	20 Apr 2026 02:16
Last Modified:	20 Apr 2026 02:16
URII:	http://shdl.mmu.edu.my/id/eprint/15753

Downloads

Edit (login required)