Improper activities appear nowadays for the human (i.e.) falling without aware, and numerous techniques had been developed to reduce them. In this essay, critically analysis of the various proposed methodologies by comparing their strength and their weakness. The complexity and diversity of actions make it difficult to recognise them. The conventional (CCTV) method is ineffective and expensive for monitoring patient activities by using sensor based is also difficult due to drained battery life. So we need real time system for activity recognition with more efficiency and accuracy to avoid people from morality problems or it may lead to causes to major injuries. By comparing various algorithm Support vector machine (SVM) is a discriminative classifier belonging tosupervised learning. Recurrent neural network (RNN) is one of the concepts in deepneural network. The main intention of the RNN is to minimize the preprocessing. This application will seem like visual imagery analysis. The convolutional neural network (CNN) it is more cost expensive compared wearable and ambience based. Diffusion Convolutional Neural Network (DCRNN) is the branch of artificial intelligence. The neural network was trained by DCRNN algorithm. The goal of the to develop a Diffusion Convolutional Recurrent Neural network-based system for automatically detecting abnormal humanactivity. and compression of the video in real times it proves to be more effective than otherclassification algorithm.