Structured Pose-Conditioned Flow Matching for Generative 5G CSI Augmentation

2026-09-24

AI Summary

This study addresses the prohibitive cost of large-scale synchronized CSI-pose data acquisition for 5G-based human pose recognition by proposing the StructFlow-HPR framework. The method integrates a reconstruction-preserving autoencoder with a pose-conditioned Transformer, leveraging flow matching and ordinary differential equation sampling to generate augmented CSI data while strictly preserving the receiver-frequency topological structure. Experimental results demonstrate that this framework synthesizes high-fidelity, pose-aligned CSI samples in real-world 5G scenarios, substantially enhancing pose recognition performance under few-shot conditions.

Abstract

With the growing demand for privacy-preserving and occlusion-resilient human pose recognition (HPR), 5G channel state information (CSI) offers a promising contactless sensing modality by integrating communication and sensing capabilities. However, collecting large-scale synchronized CSI-pose pairs remains costly in practical 5G systems. To address this limitation, we propose StructFlow-HPR, a structured pose-conditioned flow matching framework for generative CSI augmentation. StructFlow-HPR learns a continuous latent transport process from Gaussian noise to real CSI representations under pose guidance, while preserving the receiver-frequency topology of CSI through a reconstruction-preserving autoencoder. A pose-conditioned Transformer is further designed to model the latent velocity field and generate pose-aligned CSI samples via ordinary differential equation sampling. Experiments on real-world 5G sensing data show that StructFlow-HPR can produce realistic CSI-pose pairs and improve downstream HPR performance under limited-data conditions.

Problem

• Human Pose Recognition
• 5G CSI
• Data Augmentation
• Contactless Sensing
• Limited Data

Innovation

• Flow Matching
• 5G CSI Augmentation
• Human Pose Recognition
• Pose-Conditioned Transformer
• Generative Model