Human Action Recognition: A Comprehensive Survey of Multimodal Advances, Challenges, and Emerging Directions

Abstract

Human Action Recognition (HAR) has become a pivotal field within computer vision and machine learning, with transformative applications in surveillance, healthcare, human-computer interaction, and sports analytics. Despite notable advances, a persistent gap remains between benchmark-driven performance and real-world deployment, particularly regarding cross-subject generalization, fine-grained action understanding, computational scalability, and privacy preservation. This survey provides a systematic and critical review of HAR research published between 2022 and 2025, analyzing 30 peer-reviewed articles from the IEEE Xplore digital library. We trace the progression from unimodal frameworks to multimodal fusion architectures, highlighting innovations across skeleton-, sensor-, and vision-based modalities. Key architectural trends include transformer-based models, graph neural networks, and self-supervised learning, alongside domain-specific adaptations in healthcare and sports. We also examine methodological shifts toward lightweight, privacy-aware, and generalizable systems. By synthesizing these developments, the review outlines emerging research directions and highlights priorities such as robust evaluation protocols, ethical safeguards, and deployment-ready HAR solutions, thereby guiding future work in the field.