CIDER: Enabling Robustness-Power Tradeoffs on a Computational Eyeglass.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Human-Computer Interaction

Background:

• The human eye provides insights into health and cognition.
• Continuous eye tracking in natural environments is challenging due to illumination variability and high-rate data processing.
• Existing wearable eye trackers struggle with power consumption and robustness across diverse lighting conditions.

Purpose of the Study:

• To develop a wearable eye tracker with adaptive capabilities for varying illumination.
• To address power consumption and accuracy challenges in real-world eye tracking.
• To enable continuous, non-intrusive monitoring of eye parameters.

Main Methods:

• Proposed a novel staged architecture for adaptive eye tracking.
• Implemented a low-power mode for indoor settings using a Search-Refine controller.
• Developed a robust mode for outdoor sunlight conditions.
• Focused on holistic design addressing power, accuracy, and user-effortless training.

Main Results:

• Achieved pupil center estimation error below two pixels (0.6°).
• Achieved pupil diameter estimation error of one pixel (0.22mm).
• Demonstrated low power consumption: ~7mW at 4Hz and ~32mW at 250Hz.

Conclusions:

• The CIDER system offers a power-efficient and robust solution for wearable eye tracking.
• The adaptive architecture successfully handles diverse real-world illumination conditions.
• Enables continuous eye parameter monitoring for applications in health, attention, and decision-making research.