Research & Papers

Researchers slash gaze-tracking hardware needs in half

New method uses just one camera and light source for binocular gaze tracking

Deep Dive

A team led by Tongbing Huang has developed a novel approach to binocular gaze estimation that defies conventional wisdom requiring two light sources. Their method, detailed in arXiv:2607.05473, uses a single camera and single light source by creating a 'virtual light source' positioned symmetrically to the real one.

The innovation lies in estimating a 'virtual glint' from the captured image and using polynomial regression with a new normalization factor. While performance degrades slightly compared to dual-light systems, the approach proves practical for mobile scenarios where component reduction is critical. The work was presented at VSIP 2019 and published in ACM's proceedings.

Key Points
  • Traditional gaze tracking requires 1 camera + 2 light sources, this method uses 1 camera + 1 light source
  • Introduces 'virtual light source' concept to simulate second light mathematically
  • Published in VSIP '19 proceedings after acceptance, with 4 figures and 1 table

Why It Matters

Could enable cheaper, more compact gaze tracking for mobile devices and AR/VR systems

📬 Get the top 10 AI stories daily