Comparative Analysis of Kinect-Based and Oculus-Based Gaze Region Estimation Methods in a Driving Simulator

Area of Science:

• Human-Computer Interaction
• Automotive Engineering
• Cognitive Psychology

Background:

• Driver's gaze is critical for understanding attention and performance in driving research.
• Integrating gaze data into driving simulators enhances study capabilities by linking gaze patterns to driver characteristics and outcomes.

Purpose of the Study:

• To present and evaluate two gaze region estimation modules for driving simulators.
• To compare the effectiveness of Kinect and Oculus Rift devices for gaze estimation.
• To analyze the relationship between gaze displacement and head movement for accurate gaze tracking.

Main Methods:

• Developed two gaze region estimation modules for a driving simulator: one using a 3D Kinect device and another using an Oculus Rift virtual reality headset.
• Implemented and compared four gaze estimation methods, including point-based and classifier-based approaches (MLP, SVM).
• Conducted experiments with 12 participants driving in a simulated scenario using both a large screen and the Oculus Rift display.

Main Results:

• The Oculus Rift device significantly outperformed the Kinect device for gaze estimation.
• The highest performing gaze region estimation method, utilizing the Oculus Rift, achieved an accuracy of 97.94%.
• The Oculus Rift module effectively identifies the driver's gaze region within a seven-region division of the driving scene.

Conclusions:

• The Oculus Rift-based gaze estimation module offers a highly accurate and effective solution for driving simulators.
• Integrating this module enriches simulator data, enabling multimodal driving performance analysis.
• The virtual reality experience provided by Oculus Rift enhances immersion and realism in driving simulations.