P3ID — Identification across environments
Privacy-preserving person identification with IR-UWB radar and transfer learning.
The problem
Wireless person identification avoids the need for camera images, but signal propagation changes with the surrounding environment. As a result, a model trained in one environment may perform poorly in another.
The approach
We develop P3ID, a transfer learning framework for person identification using IR-UWB radar. The framework learns from signals collected in multiple environments, with a training objective that encourages representations less dependent on the environment.
Results
The framework improves person identification accuracy by 18.06% over state-of-the-art methods.
Publication
P3ID: A Privacy-Preserving Person Identification Framework Towards Multi-Environments Based on Transfer Learning. IEEE Transactions on Mobile Computing, vol. 24, no. 1, pp. 102–116, Jan. 2025. DOI.