Wi-ZFC: A Cross-Domain Gesture Recognition System Based on DFS Zero-Frequency Circles
Z. Jane Wang, Ying Liu, Zhiyong Tao · IEEE Communications Letters · 2025
Gesture recognition using Channel State Information (CSI) is susceptible to domain variations such as environment, location, user, and orientation, which limits the generalization capability of Integrated Sensing and Communication (ISAC) systems. We propose Wi-ZFC, a domain-independent gesture sensing system, to address this issue. Wi-ZFC employs the Zero-Frequency Circle (ZFC) derived from the Doppler Frequency Shift (DFS) to segment the sensing area and characterize the dynamic components of gestures by intersecting the ZFCs of multiple links. Subsequently, the sensing area is defined by the actual motion distancesDδof the various dynamic components of the gesture. The Spatial Power Profile (SPP) is further proposed to refine the Doppler frequency of the gesture. The SPP is combined withDδto derive the actual traveled Distance Information (DI) of the gesture. Finally, a CNN-RNN network is employed to enable robust cross-domain gesture recognition.