TITLE:
A Review of Diffusion Model-Based Channel State Information Data Augmentation for Human Activity Recognition
AUTHORS:
Ruotong Zhu, Jiaxuan Li, Peng Su, Luyu Zhang, Bosheng Fu, Zhengjie Wang
KEYWORDS:
Channel State Information, Action Recognition, Diffusion Model, Data Augmentation, Cross-Scenario Adaptation, Generative AI
JOURNAL NAME:
Journal of Computer and Communications,
Vol.14 No.4,
April
24,
2026
ABSTRACT: With the rapid development of the Internet of Things and wireless communication technologies, human activity recognition based on WiFi Channel State Information (CSI) has become a core research topic in smart homes, health monitoring, and security surveillance, due to its non-invasiveness, privacy protection, and the need for no specialized equipment. CSI can accurately capture amplitude and phase perturbations of wireless channels caused by human movements, providing a reliable data basis for contactless action recognition. However, this technology faces critical challenges such as scarce labeled data, poor cross-scenario adaptability, and insufficient model generalization ability in practical applications. This paper analyzes the technical principles and research status of CSI-based action recognition, focusing on the application of diffusion models in data augmentation. By integrating generative data expansion, cross-domain feature alignment, and model robustness optimization, a complete CSI action recognition enhancement scheme is proposed. It systematically discusses how diffusion models alleviate the few-shot problem through high-fidelity sample generation and improve cross-scenario recognition performance via domain adaptive learning, providing theoretical support and technical references for the practical promotion of CSI action recognition technology.