Suggestion new monitoring system by depending on the human activity recognition videos

Abstract

Modern home monitoring system techniques, such as motion detection technology and home camera system intrusion warning, are said to be insufficient, especially to meet the needs of whole automation with flaws such as needing human interaction. We suggest a substitute system, a human activity recognition (HAR) method based on the video, and a combination of long short-term memory (LSTM) and convolution neural networks (CNN) algorithm, to address the flaws that have been found. Our suggestion doesn't need to be changed. is simply deployed utilizing just low-cost modifications to the current home security protocols, and commercially available hardware The conventional security camera may be used with ease for computer vision applications. Utilizing information on actual activity gathered by video-based sensors, we assess our strategy. By drawing Loss and Accuracy curves, we demonstrate how successful it is. Show Results demonstrate that the video-approved human activity recognition method can deliver complete home automation. The monitoring system has higher accuracy as compared to traditional camera motion detectors. The precision of the system may be improved further, and we can attain for best results. (Long-term Recurrent Cnvlution Network) implementation yields result better.