The ability to spot unusual behaviour is crucial in monitoring which is used to record anomalous human behaviour without requiring much human effort, i.e., the system ability to automatically take video and to detect abnormal activity. Human fall detection is namely the detection of a person rapidly leaping down which has significant security and safety implications. This application involves explosion risk, violence detection, theft identification and it is also used to improve security in many ways. The objective is to automatically identify and mark occurrences that need human attention using computer algorithms and deep learning approaches. This research provides a framework for action detection to address the limitations of existing methods for activity identification by using STAE (Spatial Temporal Auto Encoder) model to find anomalous activity.