A Data-Driven Approach for Gaze Tracking

• Main Authors: Huang, Kevin, Khalil, Mahmoud, Luciani, Evelyn, Melesse, Daniel, Ning, Taikang
• Format: Conference Proceeding
• Language: English
• Subjects: automatic gaze tracking; Cameras; classification; Cornea; Gaze tracking; human computer interface; k-means clustering; Lenses; Light emitting diodes; Lighting; Monitoring
• ISSN: 2164-5221
• DOI: 10.1109/ICSP.2018.8652292

Gaze tracking presents an intuitive interface for technology in today's society, with its application focus in controlling electronic devices. This paper concentrates on the design and application of an automatic gaze tracking system utilizing commodity equipment. Compared to preceding low-cost methods, the proposed method is significantly simpler, lowering the barrier of entry for this type of device, and can potentially afford more accurate tracking. Through the careful placement of the infrared (IR) light-emitting-diodes (LEDs) on the monitor and coaxially to the optical axis of the camera, the pupil was illuminated and reference glints became visible on the cornea. These glints were captured by a camera capable of detecting IR light, and were used to determine the users line of sight relative to the monitor. A linear model was used to address the horizontal and vertical components of the glints in the users eye and match them to the corresponding location point on the monitor. K-means clustering was utilized to classify the separate gaze regions with promising results.