Human Action Recognition for Disaster Detection

In this paper, in case of a panic condition in public indoor space due to fire, earthquake, terror, etc., we propose a disaster detection system that detects disaster quickly by analyzing people's behavior using deep learning model. Four types of human behavior were collected as acceleration sensor data: walking, running, going to eat after class, and running away from the appearance of robbers. The classification of four human behavior data using the proposed model showed high accuracy of 99.09%.