Design of the Challenging Behavior Monitoring System for Children with Developmental Disabilities that Combines ConvLSTM and Skeleton Keypoint

This paper proposes a system to monitor the challenging behaviors of children with developmental disabilities by utilizing the method combining ConvLSTM and Skeleton Keypoint information. The system consists of a behavior classification module, a data storage and analysis module, and an intervention module. By utilizing ConvLSTM and MoveNet, the proposed system accurately classifies challenging behaviors into various types, helping early detection and intervention. Experimental results demonstrate the effectiveness of the proposed system. By learning ConvLSTM on a challenging behavior dataset, the system achieves high accuracy even when classified into segmented types. In addition, it can be confirmed that LSTM is superior at processing keypoint information compared to other machine learning algorithms such as SVM and KNN. The proposed system has important implications for the field of developmental disability and behavioral analysis. It reduces the burden on caregivers by enabling continuous monitoring and early intervention. And Second, through intervention through Positive Behavior Support, a positive environment for children with developmental disabilities can be established.