Analytical Model for Spatial Resolution Characterization of Keystroke Recognition Using WiFi Sensing

WiFi-based indoor motion and gesture detection techniques, commonly known as WiFi sensing, leverage Channel State Information (CSI) extracted from WiFi signals and have found increasing applications in domains such as home automation. Previous works have demonstrated that WiFi-based keystroke detection is constrained by the inherent wavelength of the 5 GHz WiFi signal, achieving high accuracy only under conditions of high spatial resolution. In this letter, we perform bistatic Radar Cross-Section (RCS) measurements of the human hand during a keystroke event. Then we propose a novel theoretical model that correlates CSI sample readings with the received signal power, which is subsequently utilised for classification. We also utilise commercial off-the-shelf hardware to implement CSI-based WiFi sensing for keystroke detection in an uncontrolled indoor environment. Experimental validation of our approach reveals that the proposed model can achieve an overall detection accuracy exceeding 90%.