Wi-Fi Signals for Passive Human Identification: A Study of Three Activities

This study proposes a passive human identification system based on wireless signals. The proposed system comprises three phases; preprocessing and standardization of recorded channel state information (CSI) signals and the extraction of relevant data using principal component analysis (PCA), transformation of the signals into feature vectors, and finally construction of classification models. To evaluate the proposed approach, we collected data from ten subjects in an indoor environment while performing a set of activities. The proposed approach was tested using three different activities: walking, sitting/standing, and picking up a pen, achieving subject identification accuracy of 97.3%, 99.75%, and 65.5%, respectively. The results suggest that activity-based identification systems can serve as an effective alternative to traditional methods such as passwords and smart cards.