An Open Dataset of Pediatric Activity and Energy Expediture and Deep Learning Approach for Analysis.

Obesity in school children has increased and posed an adverse impact on the development at their adolescent phase. Physical activity and energy expenditure are important factors for diagnosing and treating obesity. To obtain such metrics, we conduct a study on the preschool children to quantitatively measure their activity using two types wearable devices (ActiGraph GT9X and our low-cost one) and metabolic instruments. In such a way, we have obtained a unique dataset that consists of motion characteristics and energy expenditure of preschool children performing daily activities. We also validate that our low-cost devices produce equivalent sensor data as ActiGraph. In our preliminary study, we use a Long Short-Term Memory model to classify activity types using raw sensor data. We expect this dataset will enable research and validation of models and algorithms to quantify pediatric activities and intensity using wearable sensors.