Automated enabling of head mounted display using gaze-depth estimation

Abstract

Recently, global companies have released OST-HMDs (Optical See-through Head Mounted Displays) for Augmented Reality. The main feature of these HMDs is that you can see virtual objects while seeing real space. However, if you do not want to see a virtual object and you want to focus on a real object, this functionality is inconvenient. In this paper, we propose a method to turn on / off the screen of HMD according to user's gaze when using an augmented reality HMD. The proposed method uses the eye-tracker attached to the mobile HMD to determine the line of sight along the distance. We put this data into a neural network to create a learning model. After the learning is completed, the gaze data is input in real time to obtain the gaze predicted distance. Through various experiments, the possibilities and limits of machine learning algorithms are grasped and suggestions for improvement are suggested.