Optimizing Velocity Thresholds for Fixation Detection in Virtual Reality Using the I-VT Algorithm

Virtual reality (VR) offers an ecologically valid and scalable platform for eye-tracking research, enabling stereoscopic 3D visualization and unrestricted participant movement. However, these immersive features introduce complexities and uncertainties when directly applying traditional 2D eye-tracking methods. The present study addresses this challenge by validating and calibrating the velocity-threshold identification algorithm for fixation detection in VR-integrated eye trackers. A VR-based fixation task was designed to capture key aspects of immersive experiences, including head rotation and vergence adjustments. Rule-based criteria were developed to optimize the algorithm's velocity threshold using features such as the number of fixations and the proportion of gaze points falling within the target. The analysis of loss functions using cross-validation identified an optimal velocity threshold of 20-35° per second across target depths ranging from 1 m to 11 m. The outcome of our study enhances the validity of eye-tracking research in VR and provides a simple, reproducible workflow for calibrating fixation detection algorithms. Improving fixation detection to better understand gaze behavior can also lead to better overall user experience supported by more effective VR interactions and optimized designs to guide user focus.