Research On Human Identification Method Using WiFi Fingerprint Features

Human identification is an important function of the security system.Commonly used identity recognition is achieved through fingerprint recognition,facial features,iris recognition and other methods,but it has the disadvantages of expensive equipment and easily stolen biological features.Since human activities will cause changes in WiFi channel state information(CSI),which has a relationship with human behavior and habits,CSI can be used as a WiFi "fingerprint" for identification.The identification method using WiFi "fingerprint" has the advantages of non-contact,low cost,etc.,showing the attractive prospects,and has become a new research direction in the academic world.This paper focuses on the topic of human identification recognition and systematically studies the personal identity recognition technology based on WiFi-CSI.The main research contents are as follows:(1)Establish a portable system with WiFi-CSI data real-time collection and extraction functions.Aiming at the problem that the existing method of using desktop computers to collect and analyze WiFi-CSI data is not easy to carry and rapid deployment,first,the structural analysis of WiFi-CSI data is carried out,and then it is developed on the embedded32-bit industrial control computer using Python language.The running CSI real-time collection program finally realized the real-time collection and extraction of CSI data on the industrial control computer,and visually analyzed the extracted results.(2)Improve the CSI data preprocessing method.In view of the problem of insufficient gait information extraction in the existing CSI data preprocessing methods,first of all,according to the phenomenon that the change trend of adjacent subcarriers in the CSI data is consistent,the missing CSI data is effectively restored through the neighborhood missing value interpolation method;Then use wavelet transform to obtain the signal components related to the human's gait in the CSI data;finally,through data normalization,the CSI data is converted into a data type acceptable by the deep learning network.(3)Build a human identification model based on CNN-Bi LSTM.Aiming at the problem of insufficient feature extraction for the method of using a simple classifier to identify CSI data,a convolutional neural network(CNN)and a bi-directional long short-term memory(Bi LSTM)parallel deep learning model.The model can be divided into two parts,one part extracts the amplitude characteristics of the gait in the CSI data through the CNN network,and the other part extracts the time sequence characteristics of the gait in the CSI data through the Bi LSTM network,and then merges these two gait features and use the Softmax function to classify,so as to realize the identification of human identities.Finally,an experimental platform was built using a router and an embedded 32-bit industrial control computer.Human identification experiment was carried out on 30 people in an indoor environment,the accurate recognition rate reached 98.7%.This results show that the method has good feasibility.