Analyze Dynamic and Static Signals in CSI Signal

With the rapid advancement of wireless sensing technologies, WiFi-based non-contact gesture recognition has emerged as a key research area in human-computer interaction. In contrast to conventional vision-based or wearable sensing methods, this approach leverages Channel State Information (CSI) to detect subtle channel variations induced by human motion, offering significant advantages in terms of privacy preservation, low deployment cost, and robustness in non-line-of-sight environments. However, CSI signals are inherently vulnerable to distortions caused by phase noise, multipath propagation, and static environmental reflections, which can severely degrade recognition accuracy. To address these challenges, this paper proposes a CSI ratio model that performs complex division on raw CSI measurements, effectively isolating dynamic gesture components from static background interference while simultaneously mitigating both amplitude and phase noise. Experimental results demonstrate that the proposed method significantly enhances trajectory reconstruction fidelity and system stability in recognizing fundamental stroke patterns and simple characters. These findings underscore the effectiveness of the CSI ratio approach in improving signal discriminability and lay a solid foundation for the practical deployment of WiFi-based sensing systems in smart homes, healthcare monitoring, and other real-world applications.