Distance Metric Learning for Large Margin Nearest Neighbor Classification
The Journal of Machine Learning Research
| Hi-index | 0.00 |
This paper introduces an on-going project with the goal of measuring and analyzing children's behavior automatically. Some key technologies, including methodologies for acquiring data, tracking a target across different cameras over time, activity recognition, interaction analysis, and behavior summarization for a target child are presented. Some encouraging results from a real system we developed in a nursery school environment are also described. As these technologies enable the content-based retrieval, comparison, and summarization of large-scale observational data, they are applicable to various purposes, such as the assessment of children's development, healthcare and diagnosis.