Beyond Physical Labels: Redefining Domains for Robust WiFi-based Gesture Recognition

2026-01-07

arXiv.org

AI Summary

This work addresses the performance degradation of WiFi-based gesture recognition in cross-environment deployments caused by distribution shifts. To overcome the limitations of conventional approaches that rely on physical-domain labels, the authors propose GesFi, a novel system that introduces a latent domain discovery mechanism to automatically uncover and implicitly align domain factors responsible for distribution shifts directly from raw WiFi signals. GesFi integrates CSI-ratio denoising, short-time Fourier transform, class-level adversarial learning, and unsupervised clustering to achieve robust cross-domain recognition without requiring physical labels. Experimental results demonstrate that GesFi improves performance by up to 78% in cross-domain tasks under both single-pair and multi-pair transceiver configurations, significantly outperforming state-of-the-art generalization methods.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Biometrics, Face, Gesture & Pose
• Application Domains: Internet of Things, Sensor Networks & Smart Cities

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Security and Privacy: Large-scale security measurements

Abstract

In this paper, we propose GesFi, a novel WiFi-based gesture recognition system that introduces WiFi latent domain mining to redefine domains directly from the data itself. GesFi first processes raw sensing data collected from WiFi receivers using CSI-ratio denoising, Short-Time Fast Fourier Transform, and visualization techniques to generate standardized input representations. It then employs class-wise adversarial learning to suppress gesture semantic and leverages unsupervised clustering to automatically uncover latent domain factors responsible for distributional shifts. These latent domains are then aligned through adversarial learning to support robust cross-domain generalization. Finally, the system is applied to the target environment for robust gesture inference. We deployed GesFi under both single-pair and multi-pair settings using commodity WiFi transceivers, and evaluated it across multiple public datasets and real-world environments. Compared to state-of-the-art baselines, GesFi achieves up to 78% and 50% performance improvements over existing adversarial methods, and consistently outperforms prior generalization approaches across most cross-domain tasks.

Problem

• WiFi-based gesture recognition
• domain generalization
• distributional shift
• cross-domain robustness
• latent domain

Innovation

• WiFi-based gesture recognition
• latent domain mining
• adversarial learning
• unsupervised clustering
• cross-domain generalization