Human-Fall detection using Video surveillance in an Indoor Domain

Abstract

This paper provides an overview on the concept of detection of human-falling using video surveillance technique. Seeing the current scenario, people are not able to devote time to assist the elderly, children and those under medical treatment; thus, there is a need of surveillance system that can help individuals to be informed in case there is an emergency. Human Fall Detection can be done using a video sequence and then analyzing it to infer if there is a human shape deformation. This technique involves the use of multiple cameras in a room to detect the human fall. It uses Deep Learning and Gaussian Mixture Model to improvise and train. Firstly, the system is trained using CNN model on previous human fall images in that area. Then, the motion is monitored using cameras and then calculations are done by syncing the video recordings to check if there is sudden change in motion. If the change is encountered, then the alarm is initiated to intimate the people concerned.