Electric Charge Variation Sensors Achieve 74.5% Eye Tracking Accuracy across 57 Real-World Scenarios with Smart Glasses

Recent advances seek to track eye movements using subtle changes in electric charge, offering a potentially seamless and low-power alternative to traditional eye-tracking technologies. Alan Magdaleno, Pietro Bonazzi, and Tommaso Polonelli, all from ETH Z ̈urich, alongside Michele Magno and colleagues, now present a comprehensive field evaluation of this contactless electrooculography technique, assessing its performance in real-world conditions. Their work demonstrates that while classification accuracy varies between individuals, averaging 74. 5% across 29 users and 100 recordings, the system remains viable for everyday use, despite being susceptible to interference from nearby electronic devices. These findings represent a crucial step towards developing practical, wearable gaze interfaces for applications ranging from human-computer interaction to augmented reality, paving the way for a new generation of intuitive and unobtrusive technologies.

Contactless Eye Tracking With Qvar Sensors

This research investigates the practical application of contactless electrooculography (EOG) using a Qvar sensor for tracking eye movements in real-world scenarios. While promising in laboratory settings, performance often diminishes in everyday environments due to individual variations and interference from electromagnetic noise. The study utilizes a Qvar sensor designed to detect electrical signals related to eye movements without direct contact. Data was collected from multiple participants in a realistic setting and analyzed using advanced techniques, including a leave-one-subject-out cross-validation method and t-distributed stochastic neighbor embedding (t-SNE) visualization.

Researchers carefully quantified the impact of noise, particularly from laptops, on signal quality. Results demonstrate significant variability in accuracy between subjects, ranging from 57% to 89%, highlighting the difficulty of creating a single model for everyone. Electromagnetic noise from devices like laptops significantly degraded signal quality, with closer proximity resulting in a weaker signal. t-SNE visualizations revealed that individual subject characteristics dominate over differences between movements. This work contributes to the field by providing a thorough evaluation of contactless EOG in a realistic environment, characterizing the impact of electromagnetic noise, and highlighting the importance of addressing individual differences in signal characteristics. These findings suggest the need for adaptive algorithms that personalize eye-tracking models to individual users. Further research is needed to develop effective noise reduction techniques and to inform the design of wearable eye-tracking systems that shield against electromagnetic interference, ultimately contributing to more natural human-computer interaction systems.

Wearable Eye-Tracking via Charge Variation Sensing

Researchers pioneered a new method for contactless eye-tracking using a charge variation (QVar) sensing system, designed for integration into wearable devices. They engineered a custom hardware module incorporating six QVar channels and utilized the ST1VAFE3BX chip from STMicroelectronics to enable in-sensor processing, including signal filtering and step detection. This compact module, measuring just 19. 5mm x 16. 5mm, allows for flexible electrode placement in wearable designs.

The system employs five sensing channels utilizing ten electrodes strategically positioned on a glasses frame to capture eye movements. Researchers carefully selected electrode materials based on their performance characteristics, using softpulse electrodes from Dätwyler for contact sensing due to their skin compatibility and low impedance, and ENIG-coated copper sheets for contactless sensing due to their reliability and capacitive coupling. Two channels used contact electrodes at the nose pads and temples, while the remaining three contactless channels utilized copper sheets around the left eye for horizontal and vertical tracking. Recordings were conducted with 29 participants across five separate sessions, tracking a circle on a screen while minimizing head movement and blinking.

Each change in the circle’s position, including up, down, left, right, and diagonal variations, was recorded with a timestamp, and blink events were captured by displaying the word “Blink”. A detailed noise analysis was performed, revealing significant measurement differences even under controlled conditions. Researchers systematically assessed the impact of nearby electronic devices, such as laptops and monitors, on signal quality, and designed a test to quantify baseline noise levels at varying distances from an idle laptop. This meticulous approach enabled the team to isolate and characterize the influence of environmental factors on the system’s performance.

Contactless Eye-Tracking Achieves 74. 5% Accuracy in Real-World

Scientists achieved a mean classification accuracy of 74. 5% using a novel contactless eye-tracking system based on electric charge variation sensing, demonstrating viable performance in real-world conditions. The research involved a field evaluation across 29 users and 100 recordings captured during everyday activities, such as working in front of a laptop, to assess the system’s limitations and robustness. Results show individual user accuracy varied between 57% and 89%, highlighting the impact of subject variability on system performance. The team measured the system’s ability to classify eye movements using a lightweight convolutional neural network deployed directly on the wearable glasses for real-time processing.

Experiments revealed a significant degradation in classification accuracy in the presence of nearby electronic noise sources, indicating environmental factors pose a challenge for reliable operation. The study confirms the potential of this contactless electrooculography system as a low-power and unobtrusive alternative to traditional skin-contact electrodes, demonstrating the feasibility of capturing horizontal and vertical eye motion signals from a short distance. This work presents one of the first demonstrations of contactless eye-tracking in everyday settings, providing concrete evidence of its practical feasibility and limitations for applications like human-computer interaction and augmented reality.

Contactless Eye Tracking, Individual Variation, Noise Impact

This study presents an in-depth field evaluation of contactless electrooculography, a technique for tracking eye movements using electric charge variation sensing. Experiments conducted with 29 users demonstrate that while the system functions in everyday settings, classification accuracy varies considerably between individuals, averaging 74. 5% but ranging from 57% to 89%. The research confirms the viability of this technology outside of controlled laboratory conditions, yet highlights the challenges posed by individual differences in signal characteristics and the presence of environmental electromagnetic noise.

Specifically, the team quantified the impact of nearby electronic devices, finding that proximity to a laptop significantly reduces signal quality. Analysis reveals that signal characteristics are strongly subject-dependent, limiting the ability to create universally applicable models for eye movement recognition. Despite these limitations, the findings provide crucial insights for developing more robust and adaptive algorithms, paving the way for improved contactless eye-tracking systems suitable for wearable human-computer interaction and augmented reality applications. Future work will likely focus on mitigating the effects of noise and developing more personalized models to enhance performance and reliability.