RFPose-GAN: Data Augmentation for RFID based 3D Human Pose Tracking

In the age of Artificial Intelligence of Things (AIoT), human pose tracking has attracted increasing interest in many fields. To address the limitations of conventional vision based pose tracking techniques, Radio Frequency (RF) based pose monitoring has been proposed in recent years. However, most of the existing RF-based approaches depend on a vision-aided multi-model learning model, which requires extensive labeled data for supervised training. Collecting such large amounts of training data is time-consuming and costly. In this paper, we address this issue by proposing a Generative Adversarial Network (GAN) based data augmentation method, termed RFPose-GAN, to generate synthesized datasets to assist the training of multi-model neural networks. Our experimental results demonstrate the efficacy of the proposed data augmentation approach on improving the performance of 3D human pose tracking when there is only a limited amount of training data.