With the widespread adoption of WiFi technology, WiFi-based human sensing has shown significant potential in healthcare applications, particularly for spinal posture assessment. Conventional diagnostic methods, such as X-ray imaging, involve radiation exposure and require specialized equipment, making them costly and unsuitable for frequent, non-clinical posture monitoring. In contrast, WiFi-based sensing offers a non-contact, non-invasive, and cost-effective method for this purpose, without posing health risks. However, WiFi signals are highly susceptible to interference in complex electromagnetic environments, which can affect assessment reliability. To address this challenge, we propose a novel data representation that integrates amplitude and phase information, along with a dynamic multiscale spatiotemporal feature fusion method to enhance feature extraction. Building on this, we develop the MSF-BiTNet framework for differentiating scoliosis-related postures using WiFi signals and validate its effectiveness and robustness. Experimental results validate the effectiveness of the proposed framework in this posture assessment task and demonstrate its high classification accuracy and outstanding generalization across multiple public datasets. Compared to conventional models, our approach significantly improves classification performance. Additionally, we conduct a comprehensive evaluation of the model under various experimental conditions.
Gao et al. (Mon,) studied this question.