Experimental Validation of Interface Aiding Human Depth Prediction on Images in terms of Usability

Abstract

The purpose of this study is to maximize the remote participation experience of disabled and elderly people who have mobility problems. We reviewed the prior art on visual interfaces from two aspects and proposed a stereo matching interface to improve their combined performance. The first aspect is hedonic quality as it relates to reality of going-out experience and usability. The second aspect is that of utility. Our visual interface is a human-assisted sensing system that enables the user to move the remote-controlled robot as intended by allowing the user to predictably recognize remote depth information. Depth recognition of an arbitrary point on the display is possible just using a planar image from the robotic perspective. In this study, performances of specific tasks resulting from use of the interface were experimentally validated. The results of the stereo matching task on the display between the target feature point and its depth index as summarized by edging showed that the slower the relative approach velocity was, the easier the stereo matching task was. However, various designs created to achieve the relative approach velocity conflicted with design constraints meant to improve usability or hedonic quality.