Analytical Evaluation of Four Artificial Learning Strategies for Eye-Gaze Prediction and Tracking with Mount-free Minimal System: A Potential Communication Medium for Locked-in Syndrome Patients

Abstract

Patients with locked-in syndrome (LiS) cannot communicate effectively with caregivers due to severe physical disabilities. Since eye movement is usually the only voluntary action preserved in most paralyzed patients, it can be used as a reliable medium for communication. The main problem addressed in this study is the lack of an effective and accurate communication system for LiS patients that can convert eye movements into meaningful interpretations. The objective of this research is to design and evaluate an eye-gaze prediction and tracking system using artificial learning techniques. One of the novelties of the system is its overall design, where a mount-less minimum system, a single-camera module, is used. Moreover, the proposed system applies Machine learning algorithms to a primary eye-gaze dataset of 13 gaze-point position values generated through video frame processing of multiple healthy users. It predicts gaze-based focus points and eye movement patterns, which are then translated into corresponding audio responses and external device control commands. A user-friendly display interface with required icons is provided based on patient needs. Experimental results show that the gradient boosting algorithm achieves the highest accuracy (approximately 99.9%), outperforming decision tree, logistic regression, and random forest models.