Deep-WiID: WiFi-Based Contactless Human Identification via Deep Learning

With the widespread popularization of commercial off-the-shelf WiFi devices, the device-free WiFi sensing has attracted attention extensively. At present, some studies have explored the feasibility of WiFi-based human identification, but existing methods are facing the problem of heavy workload and low recognition accuracy. Aiming at these issues, we propose a deep learning method, named Deep-WiID, to analyze the gait feature using Channel State Information(CSI) so as to identify persons. In Deep-WiID, the Gated Recurrent Unit is combined with average pooling to extract gait features automatically from CSI data and to identify persons, which effectively reduces the overhead of data processing than traditional manual feature extraction. Experimental results conducted on CSI data collected from different situations indicate that Deep-WiID has desirable identification accuracy and good robustness. The average identification accuracy of our model is ranging from 99.7% to 97.7% when the number of persons is from 2 to 6, and there is still a desirable performance of 92.5% in larger group of 15 persons.