2021 Volume 44 Issue 4 Pages 365-376
In this study, we reported the development of an analysis method that estimates the interest of children, using position and direction information as new measures for interpreting "child's growth" in the field of early childhood education and care. In our previous studies, we showed that childcare workers empirically refer to children's position and direction when interpreting their interests. In addition, human qualitative evaluation of children’s interest states can be quantified using Bayesian estimation. Following these results, we recorded daily childcare scenes and annotated children’s behavioral features using position and direction as well as the position of the interest target provided by the childcare workers in this study. Moreover, we conducted a comparative analysis of the behavioral likelihoods of the human interest description with the machine learning methods—HMM and LDA. Our results showed that the interest tendency of the 18 participants could be estimated using position and direction information from the recorded childcare scene.
