This research studies ways to prevent physical injury for children with special needs, or specifically children with Autism Spectrum Disorder (ASD). The prevention is achievable by monitoring child behavior in the classroom from time to time. A Critical Behavior Monitoring model was developed for this purpose. The model is integrated with a Kinect sensor (by Microsoft) to process the signal acquired for human activities recognition. Currently, the model manages to identify 17 different human activities and notify parents or teachers via SMS and/or email if any unusual or critical activities are detected (i.e. falling down or asking for help). This will ensure immediate action is taken to prevent injuries or the situation from getting worse.