Development of a wearable eye tracker

Chung , Yun Sheng (2023) Development of a wearable eye tracker. Final Year Project, UTAR.

Preview

PDF Download (2116Kb) | Preview

Abstract

Eye trackers provide valuable insights into visual perception and gaze behaviour. However, the high costs and restricted mobility of current devices available in the market limit broader applications. This study aimed to develop an affordable, wearable eye tracker for naturalistic settings. Low-cost hardware and 3D-printed structures were utilized in this study to ensure cost-effectiveness. Gaze detection was started from mounting headset adjustments, followed by fine-tuning in detecting pupil, calibration, data processing on recorded data, and data analysis for calibrated data. Performance evaluations demonstrated major improvements compared to a previous model, with over 50% enhancement in angular accuracy and precision and a 52% reduction in data loss. The prototype exhibited excellent angular accuracy (0.939°) and precision (0.097°) in controlled non-movement conditions. However, facial movements and expressions generated motion artefacts that increased variability in measurements. Similarly, head and eye tracker movements produced substantial deviations in accuracy, precision, and data loss due to motion artifacts and slippage effects. A colour detection plugin was also developed, demonstrating over 90% accuracy in bright and natural lighting environments. This device could assist colour-blind individuals through real-time feedback. This ongoing colour detection could possibly assist colour-blind individuals in their daily lives as well as enhance their learning process. Further advancements can be made by improving camera sensor quality, refining algorithms, adding supplemental lighting, and collecting user feedback. Detailed validation and testing on a larger participant group across real-world scenarios could enhance effectiveness across user demographics and various use cases.

Actions (login required)