How gaze tracking works

Gaze tracking estimates where a person is looking by analyzing images of their eyes and head. Dedicated eye trackers usually shine infrared light toward the eyes and use a camera to find the pupil and the small reflection that light makes on the cornea. Webcam-based systems work from ordinary video instead, using facial landmarks and trained software models. They are generally less precise and more affected by lighting and head movement. One condition matters for anyone focused on webcam audio: gaze tracking is a video function. It does not capture, filter or enhance speech.

From eye images to a point on the screen

Most gaze tracking systems work through a similar sequence of steps. The details differ between hardware and software, but the general pipeline looks like this:

• 1. Detect the face and eyes in each camera frame.
• 2. Locate eye features such as the pupil, the iris outline and, in infrared systems, the corneal reflection.
• 3. Estimate head position and orientation, because the same eye position means something different when the head turns.
• 4. Calculate a gaze direction from the eye features and head pose.
• 5. Map that direction to a point on the screen. This step usually relies on a calibration in which the user looks at known targets.
• 6. Smooth the result over time, because the eyes make rapid small movements that would otherwise make the estimate jitter.

Infrared eye trackers and webcam-based tracking

Approach | What it relies on | General strengths | General limitations

Infrared (pupil and corneal reflection) | Infrared illumination plus a camera tuned to it | Clear eye features that are less dependent on room lighting; usually the more precise option | Requires dedicated hardware; can be affected by glasses reflections and by the user moving out of range

Webcam-based (appearance or landmark models) | A standard visible-light camera and software that infers gaze from the eye and face image | No extra hardware; works with many existing cameras | Generally coarser estimates; more sensitive to lighting, camera resolution, distance and head movement

Calibration explains much of the difference in real-world results. It links each person's eye geometry and seating position to screen locations. If the user moves the camera, changes posture or changes seats, the calibration may no longer fit, and the system may need to be recalibrated.

Gaze tracking compared with gaze correction

In video calls, the related feature people most often meet is gaze correction, sometimes labeled eye contact. Gaze correction does not report where you are looking. It changes the video image so that your eyes appear to look into the camera while you read the screen. Like other processing features, it may be provided by the camera's own software, by the operating system or by the conferencing application. Finding out which of these provides it tells you where to turn it on or off.

Gaze interaction and accessibility

In accessibility settings, gaze tracking can serve as an input method. A common pattern is dwell selection: the user looks at a target long enough, and the system treats that as a click. Gaze input is often combined with other inputs, such as switches or voice commands. In those combined setups, the microphone becomes a working part of the interface rather than only a way to be heard on calls.

What this means for capturing speech

Gaze tracking and gaze correction process video frames only. Turning them on does not change the microphone signal. Audio features such as noise suppression, echo cancellation and automatic gain are separate controls. Depending on the product, those audio features are handled by the webcam, the operating system or the application.

There is one indirect effect worth checking. Some setups lead you to turn your head away from the microphone, for example a gaze interface on a second monitor with the webcam microphone on the first. Microphones that favor sound from the front may then pick up your voice at a lower level or with a duller tone, while omnidirectional microphones are generally less affected. To check, record a short clip while facing the camera and another while facing where the gaze interface draws your attention, then compare level and clarity on playback.

If voice commands are part of a gaze-based accessibility setup, make sure the intended microphone is selected both in the operating system and in each application that listens for speech. Then confirm that its input meter responds when you speak. Exact menu locations vary by platform and version, so check them against current documentation.

Gaps in the available evidence

No product-specific evidence was available for this guide. That means it includes no accuracy figures, hardware specifications, compatibility lists, release dates or verified menu paths. The explanations above describe general, widely documented principles of how gaze tracking works. For a specific camera, eye tracker or conferencing application, rely on the manufacturer's current documentation for capabilities and settings.