Guest Editors
Prof. Elakkiya Rajasekar
Email: elakkiya@dubai.bits-pilani.ac.in
Affiliation: Department of Computer Science, Birla Institute of Technology and Science Pilani, Dubai Campus, Dubai, United Arab Emirates
Homepage:
Research Interests: computer vision, action recognition, video understanding, multimodal learning, sign language recognition/translation, temporal transformers, self-supervised learning
Summary
Action recognition and human behavior understanding have become central to modern intelligent systems, driven by the widespread availability of video sensors and the emergence of powerful deep learning models. Recent progress in temporal transformers, self-supervised representation learning, large-scale video pretraining, and multimodal fusion has substantially improved recognition of complex activities in unconstrained environments. However, real-world deployment still faces key challenges including fine-grained action discrimination, cross-domain generalization, robustness under occlusion and viewpoint changes, learning with limited labels, privacy constraints, and efficient inference for edge devices.
This Special Issue aims to collect high-quality research that advances the theory and practice of action recognition and multimodal human behavior understanding. The scope includes novel architectures and learning paradigms for video and temporal modeling, multimodal methods combining video with audio, text, depth, or skeleton streams, and application-driven solutions in smart environments, human–computer interaction, assistive technologies, and security. We particularly encourage submissions addressing generalization, low-resource learning, interpretability, fairness, privacy-preserving pipelines, and compute-efficient deployment.
Suggested themes include (but are not limited to): video transformers and temporal modeling; self-/semi-supervised video learning; multimodal fusion for behavior analysis; skeleton/pose-based action understanding; egocentric and first-person activity recognition; fine-grained and long-term activity modeling; robustness and domain adaptation; privacy-aware action recognition; real-time and edge-efficient inference.
Keywords
action recognition, video understanding, multimodal learning, temporal modeling, video transformers, self-supervised learning, skeleton-based recognition, egocentric vision, human behavior analysis, efficient / edge inference
Login
Register
Submit a Paper
Propose a Special lssue