Abstract—This study is about the automatic engagement level
measuring system which extracts features of kids taking tests
with a desktop computer and estimates an engagement level. We
recorded 12 kids from two difference kindergartens for 5 days.
The test consists of 6 subjects with 2 sessions (levels). The
recorded RGB video data is divided into 30 second video clips
which are labeled by an expert. Cues reflecting face and head
information are extracted from video data. The cues are
aggregated for 30 seconds and used for estimating the
engagement level. We used a relevance vector classifier to
estimate an engagement level. We also analyze the data using
linear regression analysis and find valid features. The system
shows a promising performance of engagement level estimation
of kids.
Index Terms—Engagement level estimation, facial expression,
facial feature, kid, multiple intelligence.
Woo-han Yun, Dongjin Lee, Chankyu Park, and Jaehong Kim are with
Human-Robot Interaction Section, Electronics and Telecommunications
Research Institute, Daejeon, South Korea (e-mail: yochin@etri.re.kr,
robin2002@etri.re.kr, parkck@etri.re.kr, jhkim504@etri.re.kr).
Cite: Woo-Han Yun, Dongjin Lee, Chankyu Park, and Jaehong Kim, "Automatic Engagement Level Estimation of Kids in a Learning Environment," International Journal of Machine Learning and Computing vol. 5, no. 2, pp. 148-152, 2015.