Dr. Ivan Vajs, a researcher at the Innovation Center of the School of Electrical Engineering in Belgrade, presented the paper “Down-sampling effects on eye-tracking features in identifying reading difficulties in children” at the IEEE ERK 2026 international conference (Electrotechnical and Computer Science Conference), traditionally held in Portorož, Slovenia.
His co-authors on the paper are Tijana Bogdanović, Dejan Drajić, Vladislava Krsmanović, Vanja Ković, and Milica M. Janković.
The paper investigates the influence of sampling frequency – one of the key technical parameters of eye-tracking devices – on features obtained during a reading task. The data of 24 children with reading difficulties and 24 children from the control group, originally recorded at 250 Hz and then down-sampled to 200, 150, 100, 50, and 30 Hz, were analyzed. Using logistic regression with feature selection, an average classification accuracy of 88% to 90% was achieved across different frequencies.
The results show that separability between children with reading difficulties and typically developing children can be detected even at low sampling frequencies, opening the door to the use of more accessible and affordable devices. However, eye-tracking feature values change with sampling frequency, so comparing them requires careful consideration of the technical characteristics of the devices.
