The invention discloses a micro-action three-dimensional sensing method and
system based on WiFi
channel state information, and relates to the technical field of
wireless signal sensing. The method comprises the following steps: collecting CSI data, preprocessing, and outputting a CSI amplitude matrix and a
phase matrix; the
phase difference of adjacent antenna pairs is used for counteracting
common noise, a linear mapping model of displacement increment and
phase change is constructed, and displacement calculation is achieved; designing a space-time
TransFormer model, fusing CSI amplitude,
phase difference, displacement increment and
Doppler frequency shift characteristics through a
time sequence and space joint attention mechanism, and outputting an initial three-dimensional trajectory; optimizing the initial three-dimensional trajectory by using a self-
supervised learning mechanism based on physical consistency constraint; and finally, constructing a dynamic path filter, carrying out real-time filtering correction on the optimized trajectory in combination with extended Kalman filtering and Bayesian multipath fusion, and outputting a final three-dimensional trajectory, thereby realizing high-precision and low-
delay trajectory reconstruction without labeling in a complex scene.