Human Activity Recognition Based Child Monitoring System

In recent times, nuclear families with both parents working round the clock are mushrooming all over the world. As a result, children are left alone in their homes without any supervision. To address this issue, this work aims to develop a ‘Human Activity Recognition based Child Monitoring System’. This system is composed of- Faster RCNN model for human activity recognition, webcam interface for input video, and an Android application. The frames are extracted from the incoming video stream from the webcam and are fed to the activity recognition model. The model then outputs the recognized activity along with a confidence score. The parent is notified through the Android application whenever the child under surveillance performs any activity that is deemed to be harmful or inappropriate by the system. The deep learning model used for human activity recognition gives quite good results for the small set of activities considered for this work.