Single Person Identification Using Wi-Fi Signals

This is the authors' abstract. We don't add key points for this paper.

Over time, identification devices and applications have evolved. However, they still require people to perform certain actions to obtain their identification. Recently, Channel State Information (CSI) has emerged as a promising technology that resonates strongly in the evolution of identification systems. CSI allows obtaining information on the state of the transmission channel in Wi-Fi networks and offers a high granularity of data that can help the identification of a single person. In this work, we propose a new methodology for identifying a person using Wi-Fi CSI data. For this, we pre-process the received signal to eliminate unwanted propagation effects, such as noise and outliers, and also to smooth the signal. Then we use the signal amplitude of different Wi-Fi subcarriers as features for machine learning models to perform a person identification. Our proposal was validated with a CSI dataset collected from different participants performing diverse activities. The obtained results confirm the applicability of the proposed methodology as we obtained 92% average precision for identifying different people.