This study presents an optimized, cost-effective radar sensor system for real-time, and contactless driver drowsiness detection. The system consist on a 24 GHz frequency-modulated continuous Wave radar module integrated behind the central rearview mirror for highly sensitive analysis of driver micromovements, including head tilts, and facial dynamics. The core innovation lies in the edge-deployed sensor-specific adaptation, where a custom CNN-LSTM pipeline processes complex radar returns to estimate eye opening and mouth opening metrics without visual data. This pipeline enhances the system's ability to isolate subtle motion characteristics—such as range and velocity—improving the sensing fidelity crucial for fatigue monitoring. Deployed on a resource-constrained Raspberry Pi platform, the model demonstrates robust integration and efficiency. Experimental validation on 20 subjects shows excellent performance, achieving up to 98.1% accuracy for eye blinking, and 97.9% for spontaneous yawning. This radar-based approach advances automotive sensing by providing a privacy-preserving, high-performance solution for safety-critical applications.
The engineering and experimental findings demonstrate that the fusion strategy developed in this work provides a practical and transferable solution for next-generation autonomous vehicle perception in real-world deployments.
Shibai Li, Shunxin Li· The 2026 International Confe...· 0 citations
Drowsiness-related traffic accidents are a major challenge in ensuring road safety in the whole world. One of the biggest challenges is to develop a reliable real-time driver monitoring system that works even under adverse conditions, i.e., low illumination and partial occlusion of the face. A number of existing methods have issues in scaling, cost, and real-world performance. This paper proposes a real-time driver drowsiness detection and accident prevention framework that integrates quantum machine learning in addition to the conventional learning methods. The underlying system looks at the facial video streams to detect various behavioral cues such as eye closure duration, blink frequency, yawning, and head movements. The proposed quantum-enhanced learning provides efficient feature representation and quick inference that is suitable for a real-time application. On detecting drowsiness, the system provides timely audio and visual alerts to inform the driver. Simulation results demonstrate state-of-the-art detection accuracy with very low latency, thereby establishing the efficacy and practicality of the proposed framework.
Karna Vishnu Vardhana Reddy, Kojjarapu Akhila, Yalla V V V Satya Krishna Varaprasad et al.· 2026 International Conferenc...· 0 citations
Driver drowsiness detection is a key component of modern Advanced Driver Assistance Systems (ADAS), aiming to enhance road safety through timely identification of reduced driver alertness. In this work, we present a comparative evaluation of several state-of-the-art deep learning models for drowsiness detection, including convolutional and transformer-based architectures, across multiple public datasets and a custom-compiled dataset. Our analysis focuses on cross-dataset generalization, lightweight preprocessing strategies, and suitability for on-device deployment. Experimental results reveal significant performance variability across datasets, indicating limited generalization beyond the original training domains. We further show that targeted preprocessing, including face detection and Contrast Limited Adaptive Histogram Equalization (CLAHE), consistently improves model performance without increasing training complexity. Although LiteRT conversion was explored to enable mobile deployment, inconsistent performance and incomplete model support stemming from the lack of a unified and reliable conversion framework, limited its effectiveness. These findings highlight key challenges in deploying robust, privacy-preserving, and real-time drowsiness detection systems on resource-constrained devices.
Zenon Lamprou, Georgia Christodoulou, Konstantinos Avgerinakis· 2026 ITU Kaleidoscope - AI a...· 0 citations
ABSTRACT:
Driver drowsiness is one of the primary catalysts for highway accidents globally, resulting in significant loss of life and financial damage. Traditional fatigue detection systems often rely on invasive sensors or coarse vehicular telemetry that fail to capture sudden micro-sleep events. This paper introduces a non-intrusive, real-time Driver Drowsiness Detection and Alert System utilizing computer vision and lightweight deep learning models. The proposed system captures live video streams, extracts 68 facial landmark coordinates using dlib's specialized shape predictor, and calculates temporal changes in the Eye Aspect Ratio (EAR) to determine eye-closure rates. To handle varied environmental conditions and visual occlusions, a Convolutional Neural Network (CNN) is integrated to analyze spatial eye states. Upon identifying sustained eye closure exceeding a safe threshold, the system triggers localized audio alarms and dispatches emergency SMS alerts containing real-time GPS metadata via the Twilio API. Operating seamlessly on lightweight client environments, this system balances high predictive accuracy (96.4%) with low latency, providing an affordable, robust solution to enhance smart vehicular safety.
Keywords:
Driver Drowsiness Detection, Computer Vision, Deep Learning, Convolutional Neural Networks (CNN), Eye Aspect Ratio (EAR), dlib, OpenCV, Real-time Alert System, Smart Vehicle Safety, Twilio API.
P. Priya, S. Kumar· International Scientific Jou...· 0 citations
Driver drowsiness is one of the leading causes of road accidents, highlighting the need for accurate and real-time monitoring systems to improve road safety. This paper presents an intelligent driver drowsiness detection framework that combines computer vision and machine learning techniques to classify driver states as alert, drowsy, or sleepy. The proposed system continuously captures facial images using a frontal camera and analyzes behavioral indicators such as Eye Aspect Ratio (EAR) and Mouth Opening Ratio (MOR) through OpenCV and Dlib. Deep features extracted from facial images are processed using Convolutional Neural Networks (CNN) and Support Vector Machines (SVM), while a weighted ensemble fusion strategy integrates the predictions of both models to enhance classification accuracy and robustness. To improve detection reliability, an additional low-angle camera is employed to monitor the driver’s face when head movements or downward postures reduce the visibility of the frontal camera. Experimental results demonstrate that the proposed CNN–SVM fusion model provides accurate and reliable driver state classification, making it suitable for real-time driver monitoring applications. The proposed framework offers a practical and scalable solution for reducing fatigue-related road accidents and improving intelligent transportation safety.
Priyanka Ghosh, Garima Bishnu, Madhumanti Mallick et al.· International Conference Com...· 0 citations
Driver drowsiness is among the main causes of road accidents worldwide and can lead to dreadful injuries. Early fatigue detection significantly helps to avoid crashes and enhance road safety. Conventional driver monitoring devices are typically designed to monitor behavior, such as eye blinking and yawning, or physiological indicators, such as EEG and heart rate. Moreover, single-modal systems usually experience lower accuracy because of environmental fluctuations, including lighting changes, occlusions and variations in camera angles. In this study, a hybrid CNN-EAR-based real-time driver drowsiness detection system that combines machine learning-based analyses of faces with geometrical monitoring using eye movement is presented to establish the reliability of the detection. The network utilized is that of a convolutional neural network (CNN) that identifies and classifies areas of faces and eyes under different environmental conditions. Moreover, the eye aspect ratio (EAR) method is a continuous method for assessing eye openness to identify the frequency of blink and extended eye closure. Information at the decision level of a fusion mechanism is used to combine CNN prediction and EAR threshold values to minimize false positives and enhance detection strength. Through experimental analysis, one can prove that the hybrid system is more accurate, shows better performance in terms of precision–recall, and exhibits a faster inference speed than single CNN or EAR-based methods do. The suggested framework is simple and affordable and can be incorporated into real-time intelligent driver assistance systems (IDASs). The solutions that will be implemented in the future include night vision assistance by using infrared cameras, estimation of head pose, and multimodal fatigue detection to make it more dependable.
S. Venkatesh, M. Chithra· 2026 4th International Confe...· 0 citations