Gaze Tracking Systems in Human-Computer Interaction

Gaze tracking systems capture the user’s eye movements and translate them into input signals for computers, mobile devices or immersive environments. Modern systems range from remote desktop cameras and head-mounted displays to built-in smartphone sensors and augmented-reality glasses. They rely on active illumination and high-speed imaging to detect the pupil centre and corneal reflections, and apply machine-vision algorithms to estimate point-of-gaze on a display. Advances in machine learning have improved robustness under variable lighting, head motion and partial occlusion, while low-cost hardware has brought gaze tracking into mainstream devices. Applications span accessibility aids for users with motor impairments, usability testing in web and software interfaces, hands-free control in automotive and aviation, and attention mapping in marketing. Integration with virtual and augmented realities is opening new interaction paradigms, enabling foveated rendering and gaze-driven analytics. Key challenges remain in ensuring sub-degree accuracy, minimising calibration effort, and preserving privacy. As gaze tracking becomes ubiquitous, its global significance lies in making human-computer interaction more natural, efficient and inclusive.

Research from Nature Portfolio

Recent studies have demonstrated that standard smartphone cameras, when coupled with carefully trained machine-learning models, can achieve gaze-tracking accuracy comparable to specialised eye-tracking hardware at a fraction of the cost. By analysing video streams from the front-facing camera and extracting eye region features, an end-to-end system predicts gaze position on the device screen with an average error below one degree of visual angle. Large-scale user trials have validated key oculomotor paradigms such as fixation stability, saccade amplitude and saliency mapping during image viewing. Furthermore, smartphone-based gaze data have been shown to detect reading comprehension difficulties in real time, pointing to applications in education and healthcare. This work highlights a pathway for scaling gaze-based interfaces to millions of users without additional sensors, paving the way for broad adoption in consumer, accessibility and research contexts.

Research from all publishers

Investigations into remote eye-tracker performance under naturalistic conditions reveal that head and body movements can significantly degrade data quality. Studies comparing several commercial systems found systematic offsets and increased data loss when participants adopt non-optimal poses, emphasising the need for rigorous device evaluation before deployment in unrestrained settings. Complementing this, work on head-worn trackers has quantified the impact of device slippage during speech, facial expression and rapid motion. Some wearable systems maintain stable gaze estimation despite minor movements, whereas others exhibit several degrees of drift, underscoring the importance of robust mechanical design and real-time drift correction. In parallel, the introduction of a deep-learning-based pupil segmentation framework has improved gaze estimation in low-light and artifact-prone environments. By using a convolutional network to segment the pupil region and estimate its contour with high confidence, this method achieves sub-degree accuracy at high frame rates and generalises across diverse datasets. These complementary efforts advance the reliability and versatility of gaze tracking for field deployment and interactive applications.

Gaze Tracking Systems in Human-Computer Interaction publication trend

The graph below shows the total number of articles in gaze tracking systems in human-computer interaction across all publications each year (not limited to Nature Index journals).

Technical terms

Calibration: The process of mapping raw eye-tracker measurements to screen coordinates by asking the user to fixate on known targets during setup.

Fixation: A period during which the eye remains relatively still, allowing detailed visual processing of a target region.

Saccade: A rapid eye movement that shifts the line of sight between fixation points.

Pupil Centre–Corneal Reflection (PCCR): A gaze-estimation technique that computes the vector between the pupil’s centre and one or more glints created by infrared illumination.

Gaze Estimation Algorithm: A computational method that translates eye-camera measurements into a predicted point of gaze on a display or scene.

References

1. Accelerating eye movement research via accurate and affordable smartphone eye tracking. Nature Communications (2020).
2. What to expect from your remote eye-tracker when participants are unrestrained. Behavior Research Methods (2017).
3. The impact of slippage on the data quality of head-worn eye trackers. Behavior Research Methods (2020).
4. DeepVOG: Open-source pupil segmentation and gaze estimation in neuroscience using deep learning. Journal of Neuroscience Methods (2019).