Semi-Supervised GAN with Time-Frequency Attention Mechanism for Human Activity Recognition Based on WBAN
Wei-Tao Cheng, Yu Shao, Long-Xin Li · 2025
Using traditional deep learning methods for human activity recognition (HAR) requires the support of a large amount of labeled data, however, it is difficult to obtain enough labeled data in the field of human activity recognition based on wireless body area networks (WBAN). To address this problem, we propose a semi-supervised learning model based on generative adversarial network (SGAN), which adopts the time-frequency attention mechanism (TFA) to realize the characterization of the importance of task-related features, and adds information entropy to improve the classification ability of the model. The experimental results demonstrate that the model successfully solves the problem of insufficient recognition accuracy under small sample conditions.