EHTask: Recognizing User Tasks From Eye and Head Movements in Immersive Virtual Reality.

Area of Science:

• Human-Computer Interaction
• Virtual Reality
• Cognitive Science

Background:

• Understanding visual attention in virtual reality (VR) is vital for applications like gaze prediction and rendering.
• Prior research often focused on single VR tasks and 2D environments, neglecting task-specific differences and eye-head coordination.

Purpose of the Study:

• To analyze human eye and head movement differences across distinct VR tasks.
• To develop and validate a novel method for recognizing user tasks in VR based on eye and head movements.

Main Methods:

• Collected eye and head movement data from 30 participants across four tasks (Free viewing, Visual search, Saliency, Track) in 360-degree VR videos.
• Analyzed movement patterns including fixation duration, saccade amplitude, head rotation velocity, and eye-head coordination.
• Developed EHTask, a learning-based method utilizing combined eye and head movement data for task recognition.

Main Results:

• Significant differences in eye and head movement patterns were observed across the four VR tasks.
• EHTask achieved 84.4% accuracy on the collected dataset, outperforming 2D-based methods (62.8%).
• EHTask demonstrated superior performance on a real-world dataset with 61.9% accuracy compared to 44.1% for existing methods.

Conclusions:

• Human visual attention exhibits task-specific characteristics in VR, reflected in distinct eye-head movement dynamics.
• The EHTask method offers a significant advancement in recognizing user tasks within VR environments.
• This research provides valuable insights for future VR applications requiring accurate user task identification.