An Extensive Survey on Machine Learning-Enabled Automated Human Action Recognition Models

Abstract

In recent times, human action recognition (HAR) is the most significant one because of its applications in numerous domains like entertainment, health, intelligent environments, and security and surveillance. A substantial amount of work was carried out on HAR and the authors have used various methods, like device free, wearable, and object tagged for recognizing the activities of humans. The most promising technology to assist elder people’s life is sensor-based HAR, which has enabled massive efficiency in human-centric applications. Studying HAR exhibits that the authors were interested in the day-to-day actions of humans. This paper offers a brief survey of recently developed HAR models available in the literature. The survey begins with a general architectural diagram of HAR system, along with a discussion of major modules. In addition, the design issues associated with the HAR system are elaborated on in detail. Besides, we offer an extensive survey of existing HAR models with their objectives, novelty, merits, and drawbacks. Moreover, a brief result analysis of reviewed approaches is performed. Finally, some open research issues and future scope of the work are discussed in the domain of HAR.