Passive Presence Detection Algorithm for Wi-Fi Sensing

In this paper, we derive a signal processing algorithm that enables a Wi-Fi station to passively detect the presence of a potential user in its vicinity. It is assumed that the potential user either doesn't carry a Wi-Fi device or, if it does, that its device does not participate in the detection procedure. Passive presence detection is performed by the station by means of tracking over time channel estimates obtained with packets transmitted by one or more stations in the Wi-Fi network, and determining when the user presence impacts the received signals. The proposed algorithm performs binary hypothesis testing and decides if a potential user is in the vicinity of the Wi-Fi station. It uses an estimate of the dynamic channel component power as the test statistic and compares it to a predefined threshold. As formulated in the paper, to increase the detection reliability, the power of the dynamic channel component is maximized by using an optimization procedure. Experimental results obtained with off-the-shelf WiFi devices and with the proposed algorithm are presented which demonstrate the validity of the analytical formulation, as well as the feasibility of performing passive presence detection using a Wi-Fi network. In controlled residential and enterprise settings, the proposed algorithm provided a detection rate of 99.7% for a false alarm rate of less than 1%.