Застосування інтегральних моделей око-рухової системи, побудованих за даними айтрекінгу в ортогональних напрямках

• Datum: 2026
• Hauptverfasser: Павленко, Віталій, Лукашук, Денис
• Format: Artikel
• Sprache: Ukrainisch
• Veröffentlicht: Kamianets-Podilskyi National Ivan Ohiienko University 2026
• Online Zugang: https://mcm-tech.kpnu.edu.ua/article/view/354929
• Назва журналу: Mathematical and computer modelling. Series: Technical sciences

The possibility of assessing the human psychophysiological state based on mathematical modeling of the eye movement system (EMS) using experimental eye-tracking data is investigated. To describe EMS dynamics, nonlinear integral models represented by quadratic Volterra polynomials in the form of multidimensional transient characteristics were used. Experimental “input-output” data were obtained during identification experiments with step test visual stimuli, which made it possible to construct EMS models for two orthogonal directions of eye movements: horizontal (Model1) and vertical (Model2). Based on the transient characteristics of the models, two types of diagnostic feature spaces were formed: a heuristic feature space and a feature space constructed from the coefficients of wavelet decomposition. To increase the representativeness of the datasets, data augmentation was performed by adding additive Gaussian noise with levels of 1%, 3%, and 5%. The performance of psychophysiological state classification was evaluated using the probability of correct recognition (PCR) with the Bayesian classifier and the Support Vector Machine (SVM) method using the Stratified k-Fold cross-validation procedure. An exhaustive search of feature combinations made it possible to determine the most informative combinations of two and three features in the investigated spaces. The obtained results showed that dataset augmentation and the use of multidimensional feature combinations significantly improve classification accuracy. The maximum PCR values were obtained when using heuristic feature spaces for the combined dataset formed based on Model1 and Model2. The results confirm the effectiveness of applying integral EMS models constructed from eye-tracking data in orthogonal directions in intelligent information systems for assessing the human psychophysiological state based on machine learning methods.