Using Natural Head Movements to Continually Calibrate EOG Signals.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Human-Computer Interaction

Background:

• Electrooculography (EOG) offers unobtrusive, long-term eye tracking but requires frequent recalibration due to varying sensor conditions.
• Existing EOG systems face challenges with maintaining accurate gaze estimation over time without external calibration references.

Purpose of the Study:

• To develop an automatic, continuous calibration method for EOG eye tracking.
• To enable robust gaze estimation in free-living conditions without manual recalibration.
• To fuse EOG signals with head movement data for improved accuracy.

Main Methods:

• Proposed a novel method fusing EOG signals with gyroscopic head movement measurements.
• Utilized recursive inference on a hidden Markov model to account for rotational degrees-of-freedom and uncertainties.
• Enabled continual calibration using natural eye and head movements during vestibulo-ocular reflex (VOR) activation.

Main Results:

• Achieved an average gaze estimation deviation of 3.54° compared to a standard video-based eye tracker.
• Demonstrated accuracy comparable to state-of-the-art mobile video eye trackers.
• Minimized the impact of sensor noise through signal fusion and advanced modeling.

Conclusions:

• The proposed method offers a viable solution for continuous, automatic EOG calibration.
• This approach enhances the feasibility of unobtrusive, long-duration eye tracking in real-world environments.
• Future research will focus on automatic detection of VOR in free-living scenarios.