Gaze dynamics reveal age-related physiological patterns across driving events

Strle, Gregor (author), Sodnik, Jaka (author), Stojmenova Pečečnik, Kristina (author)

Discriminating between multiple driving events using physiological signals remains challenging. This study examined whether gaze dynamics and pupil responses could distinguish multiple driving events and reveal age-related physiological patterns. Twenty-seven participants (12 young, 15 old) completed a simulated driving scenario featuring nine events while physiological signals were recorded. Statistical analyses revealed significant event-specific modulation of gaze dispersion metrics ((▫$\eta^2$▫ = 0.571–0.582, p < 0.001). Gradient boosted trees achieved 82.4% accuracy (95% CI [74.3%, 89.8%]) in classifying four events (Bicycle, DeerAlert, LowGas, StopAtGasStation) using 7 gaze-based features, with SHAP analysis identifying horizontal deviation standard deviation and median as primary discriminators. Time-series motif discovery revealed consistent gaze motif amplitudes (3.0–3.4 z-scores) across events, while pupil motifs showed selective enhancement during monitoring tasks (3.1–3.8 z-scores). Age-stratified analyses uncovered distinct physiological patterns: younger drivers exhibited greater gaze variability with broader scanning patterns, whereas older drivers demonstrated elevated pupil motif amplitudes (particularly during LowGas: 3.88 vs. 3.18 z-scores) alongside more constrained visual exploration. The convergence of statistical, machine learning, and motif discovery approaches establishes gaze dynamics as sufficient for multiclass event discrimination in simulated driving. These findings demonstrate that gaze-based features alone can reliably distinguish between specific driving events, while motif analysis reveals temporal dynamics and age-related patterns that warrant further investigation in larger samples, providing a foundation for interpretable driver monitoring systems.

• Language: English
• Keywords: physiological patterns, aging, driver behavior, driver monitoring systems, machine learning, time-series analysis
• Work type: Article
• Typology: 1.01 - Original Scientific Article
• Organization: FE - Faculty of Electrical Engineering
• Publication status: Published
• Publication version: Version of Record
• Year: 2026
• Number of pages: 22 str.
• Numbering: Vol. 213, art. 103827
• PID: 20.500.12556/RUL-182476
• UDC: 004.85:629.072-053.9:159.955
• ISSN on article: 1071-5819
• DOI: 10.1016/j.ijhcs.2026.103827
• COBISS.SI-ID: 277955075
• Publication date in RUL: 13.05.2026

Record is a part of a journal

• Title: International journal of human-computer studies
• Shortened title: Int. j. hum.-comput. stud.
• Publisher: Academic Press
• ISSN: 1071-5819
• COBISS.SI-ID: 3362343

Licences

• License: CC BY 4.0, Creative Commons Attribution 4.0 International
• Link: http://creativecommons.org/licenses/by/4.0/
• Description: This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.

Secondary language

• Language: Slovenian
• Keywords: fiziološki vzorci, staranje, vodenje voznika, sistemi za spremljanje voznika, strojno učenje, analiza časovnih vrst

Projects

• Funder: ARIS - Slovenian Research and Innovation Agency
• Project number: P2-0246
• Name: ICT4QoL - Informacijsko komunikacijske tehnologije za kakovostno življenje

• Funder: EC - European Commission
• Project number: 101147819
• Name: Federated cybeR-physical infrastructure for ODD cOntinuity
• Acronym: FRODDO