CSIID: WiFi-based Human Identification via Deep Learning

Abstract

With the widespread popularization of commercial off-the-shelf (COTS) WiFi devices, the device-free WiFi sensing has attracted attention extensively. However, there are only a few studies on human identification by using noncontact techniques since traditional methods are facing the problem of heavy workload and low recognition accuracy. Aiming at these issues, we propose a deep learning method, named CSIID, to analyze the gait features using Channel State Information (CSI) of COTS WiFi devices In CSIID, the convolution layers are combined with long short-term memory (LSTM) layers to extract gait features automatically from CSI data and to identify persons, which effectively reduces the need for a large amount of data preprocessing by manual feature extraction. Experimental results conducted on CSI data collected from different situations indicate that the CSIID has desirable identification accuracy. The average identification accuracy of CSIID is ranging from 97.4% to 94.8% when the number of persons is from 2 to 6.