Eye tracking technology offers invaluable insights into visual attention and cognitive processes by analyzing eye movements, gaze points, and blinks. Its applications span diverse fields, including medical rehabilitation (e.g., assisting ALS patients), market research (determining consumer preferences), and human-computer interaction (HCI) in virtual reality (VR) systems. Existing eye-tracking methods, such as scleral search coils, magnetic resonance imaging (MRI), video oculography, and electrooculography (EOG), each present limitations. Scleral coils are invasive, MRI systems are bulky and lack portability, video oculography raises privacy concerns due to camera placement, and contact-based EOG electrodes can cause discomfort and infection risks. This research addresses these limitations by proposing a novel non-contact eye-tracking system based on electrostatic induction, leveraging advancements in triboelectric nanogenerators (TENGs). TENGs offer advantages of low cost, high sensitivity, and multimode operation, making them suitable for wearable sensors. The electrostatic field generated by a charged dielectric film in a TENG interacts with moving objects, producing detectable electrical signals. This principle is adapted for eye tracking, creating a non-contact, comfortable, and portable system.

The authors review existing eye-tracking technologies, highlighting their respective strengths and weaknesses. Scleral search coils, while offering high precision, are invasive. Magnetic resonance-based systems provide high accuracy but are cumbersome and expensive. Video-based systems, though widely used, may compromise privacy. Electrooculography (EOG) uses non-invasive electrodes, but contact-based systems may cause discomfort and pose infection risks. The authors highlight the limitations of these techniques and introduce the concept of electrostatic induction as a potential solution for overcoming these challenges. The existing literature on triboelectric nanogenerators (TENGs) and their applications in sensing is discussed, setting the stage for the proposed approach.

The researchers designed a transparent, flexible electrostatic sensing interface with a triple-layer structure. This consists of a pre-charged composite dielectric bilayer (PCTFE grafted onto PDMS) and a rough-surface silver nanowire (Ag NW) electrode on a stretchable polydimethylsiloxane (PDMS) substrate. The PCTFE was chosen for its high charge storage capability and optical transparency. The rough Ag NW surface enhances the interface's capacitance, improving charge storage. The interface is pre-charged using corona poling. The system works by detecting the changes in electrostatic field induced by eye movements (oculogyria and blinks) in a non-contact manner. The signals from four channels arranged around the eye are acquired and processed using deep learning (VGG neural network) to decode eye movements and gaze direction. Material selection and optimization were guided by experimental testing and density functional theory (DFT) calculations. The fabrication process involved plasma treatment of PDMS to improve adhesion, etching of Ag NWs for a rough surface, and grafting of PCTFE onto PDMS to create the composite dielectric bilayer. The device's performance was evaluated under various conditions (bending, stretching, humidity, airflow, temperature) to assess its robustness and stability. A 3D face scan was used to analyze periocular skin movement patterns to optimize the placement of the four sensing channels in the array. The AET system was tested for its ability to track eye movements for preference analysis and as an eye-controlled input modality. Deep learning was utilized to translate the electrical signals into mouse actions (clicks, double clicks, directional movements).

The optimized triple-layer electrostatic interface achieved a high charge density of 1671.10 µC·m⁻² and a charge-keeping rate of 96.91% after 1000 non-contact operation cycles. The system demonstrated an angular resolution of 5° in detecting eye movements. The deep learning model achieved a 97% accuracy in decoding eye movements for gaze tracking and preference analysis. The eye-controlled input modality achieved 100% accuracy in translating eye movements into mouse actions, enabling a hand-free HCI system. The interface showed good stability under various environmental conditions (humidity, airflow, temperature). Specifically, the high charge density (1671.10 µC·m⁻²) of the interface, resulting from the combination of the rough Ag NW electrode and the composite dielectric bilayer (PCTFE grafted on PDMS), enabled precise and stable detection of subtle eye movements. The 3D face scan data guided the optimal placement of the four-channel array for maximum sensitivity to both vertical and horizontal eye movements, as well as blinks. The high accuracy of the deep learning model in decoding eye movements into meaningful actions confirmed the system's effectiveness. The stability tests under different environmental factors further validated its practical applicability.

The developed AET system successfully addresses the limitations of existing eye-tracking methods by offering a non-contact, comfortable, portable, and highly accurate solution. The high sensitivity and stability of the electrostatic interface, combined with the robust deep learning model, allow for real-time decoding of complex eye movements. The applications demonstrated—consumer preference analysis and eye-controlled HCI—highlight the system's versatility and potential impact. The system's ability to track eye movements even when the eyes are closed suggests further applications in sleep monitoring. The successful implementation of the AET system opens up new possibilities for applications requiring precise and comfortable eye tracking. The use of deep learning for signal processing significantly improves accuracy and reduces latency. The non-contact nature of the system enhances user comfort and avoids the potential drawbacks of contact-based methods.

This study presents a novel active eye-tracking system based on a flexible, transparent, and highly persistent electrostatic interface. The system's high sensitivity, accuracy, and stability make it suitable for various applications, including consumer preference analysis and eye-controlled human-computer interaction. Future research could explore its integration with other wearable technologies and its application in clinical settings for monitoring patients with neurological conditions.

While the system demonstrates high accuracy and stability, potential limitations include the need for calibration to individual users and the influence of external electromagnetic fields on the electrostatic readings. The current study focused on a limited number of participants. Further studies with larger and more diverse populations are needed to fully validate the system's generalizability.