Guest Editor(s)
Assoc. Prof. Dr. Anuchit Jitpattanakul
Email: anuchit.j@sci.kmutnb.ac.th
Affiliation: Department of Mathematics, King Mongkut's University of Technology North Bangkok, Bangkok, Thailand
Homepage:
Research Interests: deep learning for wearable sensor-based human activity recognition, efficient and lightweight models for edge deployment, transfer learning and domain adaptation, self-supervised learning, multimodal sensor fusion; applications in healthcare and rehabilitation
Summary
Wearable inertial and physiological sensors, together with vision-based and radar-based sensing, have made continuous human activity recognition (HAR) practical across a wide range of real-world settings, and deep learning now underpins most state-of-the-art recognition pipelines. Real deployments nevertheless remain difficult: labelled data are scarce and expensive, models degrade across users, sensor modalities and devices, and the computational budget available on wearable, embedded or edge platforms is severely constrained.
This Special Issue aims to bring together original research and comprehensive reviews that close the gap between benchmark performance and reliable real-world use of HAR, spanning wearable sensor-based, vision-based and other sensing modalities. It welcomes contributions on efficient architectures for on-device inference, learning strategies that reduce dependence on large labelled datasets, methods that generalise across subjects, hardware and modalities, and principled evaluation.
- Efficient and lightweight deep models for on-device and edge deployment, including quantisation, pruning and knowledge distillation
- Vision-based and multimodal (wearable, vision, radar) human activity recognition
- Transfer learning, domain adaptation and cross-subject, cross-device or cross-modality generalisation
- Self-supervised, semi-supervised and few-shot learning for label-scarce data
- Multimodal sensor fusion and temporal sequence modelling
- Robustness to noise, occlusion, sensor displacement and missing channels
- Explainability, uncertainty estimation, and privacy-preserving or federated learning
Keywords
human activity recognition, wearable sensors, vision-based recognition, deep learning, edge computing, transfer learning, self-supervised learning, multimodal sensor fusion
Login
Register
Submit a Paper
Propose a Special lssue