Abstarct: Nowadays, with the increase in the average age of people in the community, the issue of care for the elderly is raised. Given the advances made in intelligent care systems, the issue of intelligent video surveillance is very important. One of the applications of intelligent video surveillance is to discuss the care of the elderly and analyze their movements in order to be able to inform the family if any unpleasant events occur. In this dissertation, an applied method is used to detect the normal and abnormal movements of individuals using visual machine techniques. The proposed method is a subset of human motion analysis systems. Analysis of such systems involves several major steps. In order to analyze the movements of the events in the first step, we begin to capture the environment of the elderly and, by removing the background removal, we will identify the animated individuals in the scene. In the next step, the extraction operation is performed from the most prominent characteristics of the moving person and compression of the data is provided by a compact feature vector. In the next step, constructing and evaluating the neural network is used to learn extracted, trained, and used to test and evaluate the type of movement; The Adaboost algorithm is also used to classify k nearest neighbors to compare the accuracy of classifications. For this purpose, a databaxse of all types of human movements including standing up, sitting, taking, falling down, sitting on the chair, bending the neck and ... by different people with different appearance features and simulation reports baxsed on this databaxse and comparing it It is baxsed on similar work that allows us to identify the natural and abnormal movements of individuals using image processing techniques and neural networks. The results obtained by using the neural network method show a 97.1% accuracy in the correct determination of the movements of individuals, and comparing it with other methods indicates a high performance.