StructFlow-HPR for Pose-Conditioned 5G CSI Data Augmentation

StructFlow-HPR uses a pose-conditioned Transformer and flow matching to generate synthetic 5G channel state information (CSI) from Gaussian noise. The method employs a reconstruction-preserving autoencoder to maintain receiver-frequency topology, enabling privacy-preserving human pose recognition training.