A basic study of activity type detection and energy expenditure estimation for children and youth in daily life using 3-axis accelerometer and 3-stage cascaded artificial neural network

It is important to prevent obesity in childhood given that many obese adults have been obese since childhood. An activity monitor could provide an effective aid in preventing obesity if it records not only the calorie assessment but also activity detection to check how active a child is in daily life. The current study is for activity monitoring algorithm and we designed 3-stage cascaded artificial neural network. To develop the algorithm, we recruited 76 participants, made 3-axis accelerometer for them, and acquired activity data and calorie consumption data through them. Finally, we designed 3-stage cascaded network to classify the activities and to assess energy consumption. The 3-stage network classifies 4 activities of walking, running, stairs moving, and jumping rope with overall accuracy of 94.70%, and predicts calorie consumption with average accuracy of 81.91%, which is better than the results of the 2-stage network. Future work would include the enhancement of the network performance.