Detecting anger-provoking events with smartphones: a naturalistic driving study

Area of Science:

• * Transportation safety research
• * Affective computing
• * Human-computer interaction

Background:

• * Road safety is significantly impacted by driver anger, necessitating effective detection and intervention strategies.
• * Previous anger detection methods were primarily tested in simulated environments, requiring validation in real-world driving conditions.

Purpose of the Study:

• * To validate a driving anger detection method using a naturalistic driving study (NDS).
• * To identify common anger triggers and driving scenarios associated with driver anger.
• * To present anger-sensitive features for improved emotion detection in vehicles.

Main Methods:

• * A naturalistic driving study (NDS) involving 34 drivers over one to two weeks.
• * Data collection via smartphone application: anger-provoking events, facial expressions, and vehicle kinematics.
• * Anger-related traits assessed through questionnaires; 570 anger events collected.

Main Results:

• * The driving anger detection model achieved 93.6% accuracy and 93.4% F1 score.
• * Common anger triggers included abnormal lane-changing, being blocked, and slow driving.
• * Peak anger occurrences were observed during morning rush hours on major roadways.

Conclusions:

• * The validated anger detection model shows high efficacy in real-world driving scenarios.
• * Findings support the development of emotion-aware intelligent vehicle systems.
• * Enhanced driver emotional experience and proactive intervention systems are facilitated by this research.