SOFTWARE SOLUTIONS FOR THE RECOGNITION OF VIOLENT MOVEMENTS BY VIDEO

Abstract

This article discusses modern approaches to recognizing violent actions of a person. The recognition accuracies for these methods will also be presented. In the modern world, thanks to the development of sensory and visual technologies, systems for recognizing violent human actions have become popular. The problem of aggressiveness, including that of children, has long attracted the attention of psychologists. Among the methods using violence, a few stand out: Violence detection methods using machine learning, violence detection methods using SVM, and violence detection methods using deep learning. The most popular models are the deep belief network for unsupervised learning and the convolutional neural network for supervised learning. The purpose of this study is to consider modern software solutions for recognizing violent movements from video. In the course of the work done, an analysis was made of existing machine learning and artificial intelligence algorithms for the automatic detection of aggression.