Over the past several years, we have witnessed remarkable progress in numerous computer vision applications, particularly in human activity analysis. Human action recognition, which aims to automatically examine and recognize the actions taking place in the video, has been widely applied in many applications. This paper presents a comprehensive survey of approaches and techniques in deep learning-based human activity analysis. First, we introduce the problem definition in action recognition together with its challenges. Second, we provide a comprehensive survey of feature representation methods. Third, we categorize human activity methodologies and discuss their advantages and limitations. In particular, we divide human action recognition into three main categories according to training mechanisms, i.e., supervised learning, semi-supervised learning, and self-supervised learning. We further analyze the existing network architectures, their performance, and source code availability for each main category. Fourth, we provide a detailed analysis of the existing, publicly available datasets, including small-scale and large-scale datasets for human action recognition. Finally, we discuss some open issues and future research directions.
Companion
APSIPA Transactions on Signal and Information Processing Special Issue - Learning, Security, AIoT for Emerging Communication/Networking Systems
See the other articles that are part of this special issue.