I-AOI Algorithm to Detect Eye Fixations for Near Real-Time Processing in Virtual Reality

Eye tracking research typically detects eye fixations on areas of interest (AOIs) using a two-stage approach: the Velocity-Threshold Identification (I-VT) algorithm first classifies raw gaze data into eye fixations, which are then mapped to AOIs. While effective for post-processing, this approach may be too slow for near real-time applications. The Area-of-Interest Identification (I-AOI) algorithm offers a potential alternative by identifying eye fixations directly within AOIs, but no prior work has compared the two approaches on both output similarity and computational efficiency. This study compared I-AOI- and I-VT-based approaches in a virtual reality task where 30 participants fixated on 7 sequentially highlighted spheres arranged in a 5 × 5 grid, with gaze data sampled at 100 Hz using the HTC Vive Pro. The two approaches produced similar outputs in mean eye fixation count and scanpath length, with a small but significant difference in mean eye fixation duration. However, the I-AOI-based approach ran roughly twice as fast, supporting its use for near real-time AOI-based eye fixation detection.