Monocular free-head gaze tracking method for driving electric sickbed

Abstract

Abstract Building a free-head gaze tracking model with high accuracy, simple equipment, and not limited to wearing glasses is a challenge. In this paper, a monocular free-head gaze-tracking method based on machine learning are proposed. Two lightweight, high-precision and real-time gaze-tracking models are constructed, which can measure the 2D gaze point and 3D gaze direction respectively. In addition, we combined our gaze-tracking technology with electric sickbed to create an eye-gaze control based electric sickbed system that allows the patient to control the sickbed with their eyes. The experimental results show that the measurement errors of the two models on the MPIIGaze dataset are 4.84 cm and 4.8∘ respectively. After commissioning, user feedback has shown that this eye-gaze controlled electric sickbed system can enhance the lives of patients.