Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 02, 2018 - June 05, 2018
To improve quality of life (QOL), it is important to analyze our daily activities. Among the criteria related to QOL, time-usage studies are drawing attention in the home economics field. Time-usage analysis extracts both activities and their intentions. In this paper, we propose a method for estimating gazed objects using a mobile robot to support the acquisition of detailed time-usage measurements. The gazed objects are closely related to the activities in the time-usage analysis, for example, eating and watching TV. The proposed method first recognizes the person and the surrounding objects using the camera mounted on the robot. Next the object gazed by the person is estimated based on the face orientation. If the robot decides that the estimation from the current observation is difficult, the robot changes the view direction to reduce the difficulty. The experiments verify that the accuracy of the estimation is improved by changing view directions.
