Liu, Gang, Hao, Yanling and Zou, Yixuan (2026) CIG-MAE: Cross-modal information-guided masked autoencoder for self-supervised Wi-Fi sensing. IEEE Internet of Things Journal, 13 (16). 35614 -35624.

Preview

PDF HaoY ZouY_CIG-MAE Cross-Modal Information-Guided masked_VoR.pdf - Published Version Available under License Creative Commons Attribution. Download (3MB) | Preview

Abstract

Human action recognition (HAR) using Wi-Fi channel state information (CSI) has emerged as an attractive alternative to vision-based methods due to its ubiquity, device-agnostic nature, and inherent privacy-preserving capabilities. However, the high cost of manual annotation and the limited scale of publicly available CSI datasets restrict the performance of supervised approaches. Self-supervised learning (SSL) offers a promising avenue, but existing contrastive paradigms rely on data augmentations that conflict with the physical semantics of radio signals and require large-batch training, making them poorly suited for CSI. To overcome these challenges, we introduce a cross-modal information-guided masked autoencoder (CIG-MAE) that reconstructs both the amplitude and phase of CSI using a symmetric dual-stream architecture with a high masking ratio. Specifically, we propose an adaptive information-guided masking (AIM) strategy that dynamically allocates attention to time–frequency regions with high information density to improve learning efficiency, and we incorporate a Barlow Twins (BTs) regularizer to align cross-modal representations without negative samples. Experiments on three public datasets show that CIG-MAE consistently outperforms SOTA SSL methods and even surpasses a fully supervised baseline, demonstrating superior data efficiency, robustness, and representation generalization.

Item Type:	Article
Identifier:	10.1109/JIOT.2026.3696504
Keywords:	Modeling, Wireless fidelity, Learning (artificial intelligence), Streams, Accuracy, Human activity recognition, Training, Self-supervised learning, Convolutional neural networks, Design methodology
Subjects:	Computing > Intelligent systems Computing
Date Deposited:	05 Nov 2026
Dates:	Date Publication status 25 May 2026 Published
School, department or research centre:	School of Computing and Engineering
Keywords:	Modeling, Wireless fidelity, Learning (artificial intelligence), Streams, Accuracy, Human activity recognition, Training, Self-supervised learning, Convolutional neural networks, Design methodology
URI:	https://repository.uwl.ac.uk/id/eprint/15773

Downloads

Actions (admin access)