A Machine Learning Approach for Driver Physiological Data Analysis: Regression and Classification

Öz

Trafik kazaları, küresel çapta büyük kayıplara yol açan ciddi bir sorundur ve temelinde yorgunluk, dikkat kaybı ile uyku hali gibi insani faktörler yatar. Erken dönem ADAS sistemleri dış tehditleri ve araç dinamiklerini izleyerek daha çok reaktif (tepkisel) önlemlere odaklanırken, günümüzde proaktif (önleyici) Sürücü İzleme Sistemlerine (SİS) doğru temel bir paradigma değişimi yaşanmaktadır. Bu modern sistemler, kritik bir durum henüz ortaya çıkmadan riskleri tahmin edip önleyebilmek adına doğrudan sürücünün vücudundan alınan fizyolojik verileri kullanır. Bu çalışma, makine öğrenmesi (MÖ) modellerinin sürücü fizyolojik verilerini analiz etmede nasıl kullanıldığını incelemekte; sınıflandırma ve regresyon yaklaşımlarını karşılaştırmalı olarak ele alan kapsamlı bir literatür derlemesi sunmaktadır. Çalışmada temel fizyolojik ve davranışsal metrikler incelenmiş, çeşitli MÖ modellerinin uygulama alanları ve etkinlikleri analiz edilmiş; ayrıca veri değişkenliği, gerçek zamanlı işleme zorlukları ve sensör kısıtlamaları gibi kritik sorunlar tartışılmıştır. Derleme sonucunda, derin öğrenme yöntemlerinin artan gücü ve bu sistemlerin gerçek dünya koşullarında sağlam, genellenebilir bir şekilde çalışabilmesi için çok modlu sensör füzyonunun taşıdığı stratejik önem açıkça vurgulanmaktadır.

Anahtar Kelimeler

• Makine Öğrenmesi (MÖ)
• Derin Öğrenme (DÖ)
• Sürücü İzleme Sistemi (SİS)
• Fizyolojik Sinyaller
• ADAS

A Machine Learning Approach for Driver Physiological Data Analysis: Regression and Classification

Öz

Traffic collisions remain a devastating global issue. Human error — specifically fatigue, loss of focus, and drowsiness — plays a massive role in these tragic events. Early Advanced Driver Assistance Systems (ADAS) mainly took a reactive approach, keeping an eye on external threats and vehicle movements. However, the automotive industry is currently undergoing a massive philosophical shift toward proactive Driver Monitoring Systems (DMS). Instead of waiting for a car to swerve, these modern setups look directly at the driver's own physiological data. By doing so, they can predict and neutralize risks long before an actual emergency happens. This paper takes a deep dive into how researchers are applying machine learning (ML) models to analyze this physiological data, specifically comparing classification and regression techniques. We break down the core behavioral and physical metrics, evaluate how well different ML algorithms actually perform, and tackle the messy, real-world challenges like sensor limits and unpredictable data. Ultimately, this review points out that deep learning is rapidly taking over the field, and combining different types of sensors (multimodal fusion) is the smartest strategy for building reliable systems that work in chaotic, everyday driving conditions.

Anahtar Kelimeler

• Machine Learning (ML)
• Deep Learning (DL)
• Driver Monitoring System (DMS)
• Physiological Signals
• ADAS

Kaynakça

1. [1] NHTSA, "Driver Assistance Technologies," accessed Sep. 2025. [Online]. Available: https://www.nhtsa.gov/vehicle-safety/driver-assistance-technologies.
2. [2] KORE Wireless, "What Is a Driver-Monitoring System (DMS)?" accessed Sep. 2025.
3. [3] K. Jin, A. Rubio-Solis, R. Naik, D. Leff, J. Kinross, and G. Mylonas, "Human-Centric Cognitive State Recognition Using Physiological Signals: A Systematic Review of Machine Learning Strategies Across Application Domains," Sensors, vol. 25, no. 13, Art. 4207, 2025. doi: 10.3390/s25134207.
4. [4] ResearchGate, "Multimodal Fatigue Detection in Drivers via Physiological and Visual Signals," 2024.
5. [5] ResearchGate, "ECG-based fatigue driving detection model," 2024.
6. [6] E. T. Solovey, M. Zec, E. A. Garcia Perez, B. Reimer, and B. Mehler, "Classifying Driver Workload Using Physiological and Driving Performance Data: Two Field Studies," Proc. ACM CHI ’14, 2014.
7. [7] Y. Sun, "An In-Vehicle Physiological Signal Monitoring System for Driver Fatigue Detection," TRB, 2011.
8. [8] T. Abe, "PERCLOS-based technologies for detecting drowsiness: current evidence and future directions," SLEEP Advances, vol. 4, no. 1, zpad006, 2023. doi: 10.1093/sleepadvances/zpad006.

9. [9] ResearchGate, "Driver Fatigue Transition Prediction in Highly Automated Driving," 2020.
10. [10] SPIE, "Assessing Deep Learning Techniques for Human Driving Behavior Prediction," 2025.
11. [11] BlackBerry QNX, "What Is a Driver Monitoring System?" accessed Sep. 2025.
12. [12] S. Narayanan et al., "Estimating Markers of Driving Stress through Multimodal Physiological Monitoring," arXiv:2507.14146, 2025.
13. [13] Eventiotic, "Physiological Measurements for Determining Driver's State," 2020.
14. [14] PMC, "Driving Drowsiness Detection Using Spectral Signatures of EEG," Frontiers in Neuroscience, 2023.
15. [15] PMC, "An EOG Sensor's Ability to Detect Driver Drowsiness," 2023.
16. [16] MDPI Applied Sciences, "AI- and Deep Learning-Powered Driver Drowsiness Detection," vol. 15, no. 3, 2025.
17. [17] IRJET, "Machine Learning Based Driver Monitoring System," vol. 10, no. 12, 2023.
18. [18] Univ. of Michigan, "Predicting Driver Fatigue in Automated Driving with Explainability," 2021.
19. [19] IRJET, "Driver Drowsiness Detection System," vol. 8, no. 5, 2021.
20. [20] PMC, "Ensemble CNN to Detect Drowsy Driving with In-Vehicle Sensor Data," 2021.
21. [21] A. Shi et al., "SVM Classification of Drunk Driving Behaviour," Int. J. Environ. Res. Public Health, vol. 14, no. 1, 2017.
22. [22] ResearchGate, "SVM Classification of Drunk Driving Behaviour," 2017.
23. [23] IJARSCT, "Driver Drowsiness Detection using ML and DL," 2023.
24. [24] PMC, "Deep Learning in Physiological Signal Data: A Survey," IEEE Rev. Biomed. Eng., 2020.
25. [25] ResearchGate, "DL Model for Driver Drowsiness Detection in Low-Light Conditions," 2023.
26. [26] MDPI Applied Sciences, "CNN-Based Approach for Driver Drowsiness Detection," vol. 13, no. 13, 2023.
27. [27] TOJDAC, "Regression Model for Driving Fatigue Detection Based on Seat Pressure Distribution," vol. 8, 2018.
28. [28] T. Åkerstedt and M. Gillberg, "Subjective and objective sleepiness in the active individual," Int. J. Neuroscience, vol. 52, pp. 29–37, 1990.
29. [29] V. Putilov and O. Donskaya, "Construction and Validation of EEG Analogues of the KSS," Semantic Scholar, 2014.
30. [30] ResearchGate, "Detecting Stress During Real-World Driving Tasks Using Physiological Sensors," 2003. [31] Rheinmetall, "Sensor Fusion," 2023.
31. [32] insideAI News, "Using Machine Learning with Health Data: Challenges and Pitfalls," 2017.
32. [33] O. F. Hassan, A. F. Ibrahim, M. A. Makhlouf, and B. Hafiz, "Intelligent driver monitoring systems: a survey of drowsiness detection technologies for road safety," Artificial Intelligence Review, Springer, 2026. doi: 10.1007/s10462-026-11505-w
33. [34] A. Dontoh, S. Ivey, L. Sirbaugh, A. Danyo, and A. Aboah, "Visual Dominance and Emerging Multimodal Approaches in Distracted Driving Detection: A Review of Machine Learning Techniques," arXiv:2505.01973, 2025.
34. [35] M. S. AL-Quraishi, S. S. A. Ali, M. AL-Qurishi, T. B. Tang, and S. Elferik, "Technologies for detecting and monitoring drivers' states: A systematic review," Heliyon, vol. 10, no. 20, e39592, 2024. doi: 10.1016/j.heliyon.2024.e39592.
35. [36] L. Melders, R. Smigins, and A. Birkavs, "Recent Advances in Vehicle Driver Health Monitoring Systems," Sensors, vol. 25, no. 6, Art. 1812, 2025. doi: 10.3390/s25061812.
36. [37] Kaggle (user: gowrishankarachari), "Driver Physiological Dataset," accessed 2025. [Online]. Available: https://www.kaggle.com/datasets/gowrishankarachari/driver-physiological-dataset