Some Contemporary Approaches for Human Activity Recognition: A Survey

Abstract

Human activity recognition(HAR) in video analysis has been an prominent area of research in the field of computer vision due to its large span of applications like intelligent surveillance systems, robot learning, Human Computer Interaction(HCI), health care. An activity recognition in video involve a large amount of data to be processed making it a challenging task. In the recent years, researchers have proposed several model using deep learning with different input modalities. This paper presents an analytical survey of recent deep learning models on the basis of input modality, feature extraction, classification, dataset used and accuracy.