Iterative classification strategy for multi-resolution wireless sensing of passive targets

In this Letter, the estimation of the presence, position, and posture of a device-free target is addressed through a multi-resolution strategy that applies a virtual zoom on the information content of the channel state information measured by a single WiFi link. A series of binary classifiers are trained to estimate multiple location-based features of the target from the same measurement. Preliminary experiments point out the feasibility to estimate high-resolution features such as the target posture even in very large investigation domains, passing through the estimation of the target presence and position. A robust and ubiquitous wireless sensing is obtained with failure rates lower than 2.5% in the three considered resolution steps.