Human activity recognition(HAR) in computer vision has garnered significant attention from researchers due to its vital applications. This survey provides an extensive overview of recent approaches in HAR utilizing deep learning(DL) models. Various types of research have utilized deep learning techniques within the realm of Artificial Intelligence (AI). These investigations encompass behavior analysis, scene understanding, scene labeling, HAR, object localization, and event recognition. Among these, human activity recognition stands out as a particularly complex task and a key area of focus within video data processing research. Its applications span video surveillance systems, human-computer interaction, characterization of human behavior, and robotics. This paper's primary objective is to provide a comprehensive review of HAR, with a specific emphasis on leveraging deep learning methods. Additionally, a brief comparison is conducted, demonstrating the evolution of HAR methodologies through fusion with deep learning. The paper concludes by discussing future research directions and highlighting existing challenges in the domain of HAR.