A survey on human behavior analysis and actions recognition from videos

Abstract

Human behavior analysis using action recognition remains an active research domain of computer vision. Action prediction using Artificial Intelligence (AI) by machine learning has attracted the attention of several researchers. The presentation of human action is usually considered an important challenge. An effective representative should be robust to noise, invariant to viewpoint changes and complex scenes involving high speeds. Two main challenges in this task include how to efficiently represent spatio–temporal patterns of skeletal movements and how to learn their discriminative human features for activity classification. In this survey, we present an overview of human presence and action recognition methods used in the last years. The results of the related works are compared with our results. The performance of methods is evaluated with different human action dataset such as KTH, UCF101, UCF Sport and CAD-60.