Eye tracking via webcam: is it really possible?

We investigate new technologies related to eye movement mapping.

Eye tracking with the use of a traditional integrated webcam is possible. Technically possible, but not the choice that today we feel like adopting for our digital experience analysis projects in TSW.

Knowing the development of the sector in relation to new trends also linked to artificial intelligence and predictive models is essential, but it is also essential to guarantee a scientifically solid result of our research actions, even more so if it is a quantitative activity that investigates, thanks to psycho-physiological measurements, a series of parameters that relate to a dimension that goes beyond the declared person.

The scientific method approach is what allows us to be honest and different, in this consultancy sector called neuromarketing, but which we consider psychophysiology applied to the investigation of people’s needs and requirements (what marketing investigates).

Here, then, is an overview of this tool, which is currently at the center of interest in the scientific-technological community.

Recording eye movement from webcam: how it works

How does an eye tracker that uses a webcam work? Let’s start with the general conditions:

• First, it does not locate the pupil, but the area that also includes the iris.
• Secondly, the recording is influenced by the definition, and therefore resolution of the camera, and by the number of frames per second. To improve one of the two parameters, the other must get worse: how to understand which is the lesser evil?
• Adequate and homogeneous brightness is then necessary to allow good recognition of the face. Little light leads to a lower definition of the eye area detected and therefore a lower precision.

As far as the procedure is concerned, we advance by increasingly detailed detection steps:

1. face position
2. position of the eyes inside the face
3. left and right eye orientation
4. calibration.

Webcam eye movement detection: pros and cons

Webcam eye tracking has several potential benefits:

• remote administration, in real time
• measurement in real context (less alteration of the user’s digital experience)
• rapid execution times
• simplification of tests on smartphones and tablets
• standard camera position in relation to the screen
• tests on recent and very up-to-date devices
• greater number of users per target, reachable all over the world.

All this in the face of a major drawback: the accuracy of the detection, which limits its application to behavioral research.

In particular, the literature tells us that accuracy is incomplete for several reasons:

• extensive calibration and validation procedure
• time resolution/low speed transmission
• high sensitivity to movement
• high conditioning based on the type of lighting
• complex integration into experimental software.

Surely, thanks to the many ongoing studies, the development of new technologies and the application of machine learning, this methodology has good prospects for improving the accuracy of the detection and consequently the reliability of the resulting results.

Let’s add another critical element: the test takes place in an uncontrolled context, therefore, the researcher is not sure of any variables that differ from case to case. Trivially, you can’t be sure that a person is performing the test, because bots can simulate the required interactions.

You can try to fill this gap with a part of the questionnaire on the environmental context, of course. But it is always based on what is perceived and declared by the user, not on the external and more objective observation of the researcher.

Eye-tracking by webcam or infrared: differences

Certainly eye tracking via webcam offers considerable possibilities for the future, linked to the concept of “remote”, to increase the administration numbers of each study, obtain validations with more substantial numbers, and drastically reduce data collection times.

But, de facto, at the moment this type of solution can hardly overcome the infrared eye tracking technology which, certainly, binds to face-to-face activity, but which we at TSW continue to prefer first of all due to the greater scientific reliability, but also for a human and relational dimension linked to welcoming and listening.

The high infrared performance, especially for tests on small images and for calculating the dwell time, guarantees a very accurate detection context (around 90%), regardless of the lighting, with an ability to compensate for movements head, or other physiological changes in the eye area.

Sources:

Xiaozhi Yang, Ian Krajbich, Webcam-based online eye-tracking for behavioral research, Judgment and Decision Making, Vol. 16, No. 6, November 2021, pp. 1485–1505

Kyle Krafk, Aditya Khosla, Petr Kellnhofer, Harini Kannan, Suchendra Bhandarkar, Wojciech Matusik, Antonio Torralba, (from University of Georgia, Massachusetts Institute of Technology and MPI Informatik), Eye Tracking for Everyone, 2016 IEEE Conference on Computer Vision and Pattern Recognition

Katarzyna Wisiecka, Krzysztof Krejtz, Izabela Krejtz, Damian Sromek, Adam Cellary, Beata Lewandowska, and Andrew Duchowski. 2022. Comparison of Webcam and Remote Eye-Tracking. In 2022 Symposium on Eye-Tracking Research and Applications (ETRA ’22). Association for Computing Machinery, New York, NY, USA, Article 32, 1–7.

Kasia Wisiecka, Beata Lewandowska, Adam Cellary, Andrew T. Duchowski, Dr. Krzysztof Krejtz and Izabela Krejtz, Comparison of Webcam and Remote Eye-Tracking, NMSBA.COM