講演情報

[AI-08]Exploratory Analysis of Road Surface Vibration Effects on Occupant Heart Rate Variability Using a Driving Simulator

*MENGQI ZHENG1, Kazuya Tomiyama1, Tomoya Kitani2 (1. kitami institute of technoiogy (Japan), 2. Shizuoka University (Japan))

キーワード:

Facial Action Units、Non-contact Stress Estimation、Driving Simulator、Heart Rate Variability、Road Surface Conditions

AbstractThis study explores the feasibility of estimating mental stress of vehicle occupant due to road surface irregularities using facial expressions as a non-contact measurement method. A driving simulator experiment is conducted based on a virtual reproduction of a local road in Kitami city. Eleven participants are voluntarily participated in the experiment seated in the passenger seat under automated driving conditions to exclude cognitive load caused by driving operations. Facial video and heart rate data are recorded simultaneously during the experiment. In this study, facial expressions are analyzed using the OpenFace facial behavior analysis toolkit, which extracts Facial Action Units (AUs) defined in the Facial Action Coding System. Heart rate signals are collected using wearable sensors and normalized across participants using z-score transformation (HR_z). The driving route is segmented into event units based on road surface unevenness and curvature, which are expected to impose higher physical demands on occupants. Event-based statistical analysis and visualization were applied to evaluate localized facial and physiological responses. The results shows that upper-face AUs, particularly cheek raiser, lid tightener, and brow lowerer, exhibited higher activation levels in road sections with high mental loads. These facial responses are partially synchronized with fluctuations in normalized heart rate. Event-level AU heatmaps further reveal distinct facial activation patterns corresponding to different road conditions. These findings suggest that specific facial action units can continuously capture facial responses potentially related to occupant stress without physical contact. This study provides a fundamental evaluation framework for facial-expression-based stress estimation and demonstrates its potential applicability to in-vehicle monitoring and road environment assessment systems.