Abstract
Person Re-Identification is a key and challenging task in video surveillance.While traditional methods rely on visual data, issues like poor lighting,occlusion, and suboptimal angles often hinder performance. To address thesechallenges, we introduce WhoFi, a novel pipeline that utilizes Wi-Fi signalsfor person re-identification. Biometric features are extracted from ChannelState Information (CSI) and processed through a modular Deep Neural Network(DNN) featuring a Transformer-based encoder. The network is trained using anin-batch negative loss function to learn robust and generalizable biometricsignatures. Experiments on the NTU-Fi dataset show that our approach achievescompetitive results compared to state-of-the-art methods, confirming itseffectiveness in identifying individuals via Wi-Fi signals.