Abstract
The major drawback of Wi-Fi fingerprinting-based localization systems is the resource-heavy process of collecting fingerprint data. However, nowadays, generative models are drawing attention due to their ability to efficiently generate synthetic Wi-Fi fingerprints based on some real fingerprints collected from the target environment. Existing synthetic fingerprint generation approaches are based on generative adversarial networks or variational autoencoders. In this work, we propose a new scheme to generate synthetic Wi-Fi fingerprints using a conditional denoising diffusion probabilistic model (cDDPM). Performance evaluations with two open datasets show that the proposed method can perform well, particularly in scenarios where the number of real fingerprints is extremely limited.