Towards Private Human Sensing: A Privacy-Preserving Framework for Human Pose Reconstruction over OFDM Systems

Abstract: Wearable inertial measurement unit (IMU) sensors enable full 3D human pose reconstruction for extended reality and healthcare applications, but the motion data they capture is sensitive and must be transmitted privately and reliably over error-prone wireless channels. In this paper, we propose a privacy-preserving framework for human pose reconstruction over orthogonal frequency-division multiplexing (OFDM) systems. The framework integrates local Differential Privacy (DP) into the IMU-to-receiver transmission pipeline by adding calibrated noise to the raw IMU signals on the user’s device before transmission. At the receiver, a two-stage CNN–LSTM deep learning receiver jointly mitigates the DP noise and the wireless channel impairments to reconstruct the full-body pose. The framework is validated with realistic ray-tracing channel models on a large-scale IMU dataset, demonstrating accurate pose reconstruction under formal privacy guarantees.

Recommended citation: N. Q. Hieu, Y. F. Alem, M. A. Alsheikh, "Towards Private Human Sensing: A Privacy-Preserving Framework for Human Pose Reconstruction over OFDM Systems," IEEE Internet of Things Journal, 2026.
Download Paper