Eye-tracking-based robotic arm control system

Abstract

People with severe speech and motor impairments are unable to actively use their muscles, with the result being that they have difficulty communicating with the external world. In this study, we developed a non-invasive robot-arm control system based on eye tracking. We conducted a user-centered design process with eight commands and an intermediate real-time video transmission user interface after fully considering the spatial characteristics of the robotic arm. Additionally, we evaluated three eye-gaze point processing algorithms. Among them, density-based spatial clustering applied with a noise algorithm achieved an average accuracy of 99.3%. On this basis, we designed and conducted offline experiments, in which all five participants were able to send commands with accuracy higher than 99% for a total of 80 random commands.