Gaze estimation for social robotics: Bridging the gap: Appearance-based gaze estimation models for human-robot interaction
Robots Can Recognize Human Eye Contact Using a Standard Camera
Social robots can estimate the direction of a person’s gaze fairly accurately using only a standard camera. This does not require expensive, specialized eye-tracking equipment. This is according to research by robotics researcher Linlin Cheng on appearance-based gaze estimation for human-robot interaction.
The technology can determine whether someone is looking at a robot, for example, or at a task in front of them. This works particularly well at a distance of about one to two meters. The research also shows that a robot equipped with inexpensive cameras can recognize when someone has completed a task—for example, when the person shifts their gaze away from the task.
The technology still has limitations. Quick glances and sudden shifts in attention are more difficult to detect. As a result, the method is primarily suited for observing general attention patterns and less so for precisely tracking every eye movement.
The results could make the use of social robots in everyday life more accessible. Since only a standard camera is required, the costs and technical complexity can be lower than with systems using specialized eye trackers. This is relevant, for example, for care robots in hospitals, robots in education, and assistive robots in the workplace.
For example, a care robot could recognize whether an elderly person is looking at a pill organizer or is distracted, and adjust its behavior accordingly. In a work environment, a robot could detect from a shift in gaze that someone has completed a task.
This research thus brings robots one step closer to natural interaction in environments where people and robots work and live together. At the same time, further research is needed to ensure the technology works reliably during rapid movements and in a wide range of real-world situations.
More information about the dissertation