Toward Interest Estimation from Head Motion Using Wearable Sensors: A Case Study in Story Time for Children

Learning activities were evaluated using questionnaire survey, video, or audio. However, these methods have the following problems. First, writing on the questionnaire paper is difficult, especially for little children. Second, because the answering questionnaires was performed after the experiments were finished. They were different temporally and spatially from the scene to be evaluated. Moreover, sometimes participants has forgot the part of the contents. Finally, by recording video or audio, we can look back at each scene and evaluate them, however, video or audio analysis takes a very long time. This research aims to solve these three problems and evaluate natural reactions; first, for children of a low age group, second, including changes in the state of participants during an activity, and third, as much as possible without wasting time and effort. In this paper, during storytelling events for children, we attempted to obtain the values of acceleration and angular velocity sensors with sensors placed on the participant’s heads, and tried to estimate their motions and degree of interests. Motions were calculated using the F-value, with accuracies of 0.66 in “Sitting state”, 0.26 in “Sitting again”, 0.47 in “Wriggling”, and 0.93 in “Playing with hands”. From these results, “Playing with hands” had the highest degree of interest, with a motion recognition rate of 0.93 in F-value. Comparing the proposed method with the video evaluation later, the proposed method can obtain the evaluation result during the learning activity. Therefore, by feeding back the estimation result in real time, we can make improvements while doing activities.