Gesture Recognition for Autistic Children using Person Pose Estimation and Supervised Learning

Autism Spectrum Disorder is a neural disorder which affects the cognitive, emotional, physical and social health of the individual. Identifying the gesture of autistic children performs a vital role to prevent the meltdown and self-injury. In this research, we present a new approach to identify the gesture by identifying the poses through person pose estimation technique and use the features from person pose estimation to build a gesture classification model using supervised learning techniques. Proposed model achieved the highest accuracy for Random Forest technique with evaluation metrics of 83% for precision and 71% for recall.