Enhancement of radio frequency fingerprint data for indoor localization using diffusion model

Abstract

By gathering enough fingerprints offline to create a rich fingerprint database, the radio frequency fingerprint indoor localization method provides accuracy.Based on the diffusion model, a data augmentation technique known as FPDiffusion was presented to lower the cost of acquiring fingerprints.First, a temporal graph representation of the fingerprint sequence was created. Gaussian noise was then added to the diffusion model to complete the forward process, and a U-Net was used to complete the reverse process.The network's loss function was created with radio frequency fingerprint features in mind.