Robust Discrimination Method of the Electrooculogram Signals for Human-Computer Interaction Controlling Mobile Robot

In this paper, a robust algorithm that discriminates various eye motions from the ElectroOculoGram (EOG) signals is proposed. Previous researches that use the EOG only focused on saccadic motions or blinks. However, we cover all eye motions including double/triple blinks and left/right winks. Furthermore, we suggest a novel method, which removes noises of the EOG, to increase the robustness of the discrimination. The method is called "an ideal velocity shape algorithm" which compares the real velocity of the EOG with an ideal velocity designed under a noise free assumption. This algorithm significantly reduces the effects of the noises and thus enhances the robustness. Detected eye motions are used for a Human-Computer Interaction (HCI) between a person and a mobile robot. In the HCI, the person successfully controlled the robot for a target tracking and point stabilization.