Signal Comparison and Analysis of Various Gesture Behaviors Under Wireless Sensing Background

Channel State Information (CSI) has emerged as a critical tool for human posture analysis in the era of 6G communications and ubiquitous wireless sensing. Despite extensive research in coarse-grained activity recognition, systematic investigation of the statistical characteristics of CSI data for specific postures remains scarce. A comprehensive analysis framework is established for five representative postures—fall from sitting, fall from standing, walking, standing up and sitting down, and picking up from the ground—dividing them into fifteen distinct action groups. Four key statistical parameters, including maximum, minimum, mean, and variance, are extracted from the amplitude of CSI subcarriers. Experimental results reveal that dynamic actions exhibit significantly higher variance and a wider range of extreme values, particularly lower minimum values, due to rapid channel fluctuations, while maximum and mean values remain comparatively stable. Data visualization through two-dimensional and three-dimensional representations further elucidates these patterns. The analysis is based on a dataset comprising 9,000 CSI recordings from 30 diverse volunteers across different environments. Promising implications for future research in CSI-based sensing and human activity recognition are also discussed, paving the way for the development of lightweight algorithms in 6G edge networks for applications in smart elderly care and intelligent home monitoring.