The Effect of Interaction Using Gaze Input / Output on Smart Speaker

Abstract

In this study, we focused on the problem of existing smart speakers that natural conversation is not accomplished due to restrictions of the dialogue structure. In order to solve the problem, we proposed the method using gaze instead of wake-word based on sociological knowledge about "Initiation talk". We developed the smart speaker "Tama", that have the cameras to detect the user’s gaze and the robotic head to show the gaze for users. To evaluate the effectiveness of the proposed system, we conducted an experiment using the "Tama". For the evaluation, we compared three conditions: wake word condition, gaze detection condition , and mutual gaze condition. Experimental results showed that the usability of the system and the sense of conversation are improved when using gaze input / output.