Research On Identification Methods Based On Wireless Signals

With the development of society,people have higher and higher security requirements for privacy protection,so identity identification technology has become a research hotspot.Traditional identification technology mainly uses dentification items and identification knowledge which are easy to be lost and forged.Biometrics identification technology such as fingerprint,face image and gait makes up for the shortage of traditional identification technology.But Iris and fingerprint recognition require special equipment which are expensive.Face and gait recognition based on visual equipment rely on light,meanwhile they have visual blind corners and is easy to reveal user privacy.Because of some advantages such as low cost,wide coverage and protect user privacy,identification technology based on Wi-Fi attracts the concerns of researchers.From the perspective of indoor identity,this paper mainly conducts research from the following two aspects:(1)To improve the accuracy of identification,the paper proposes identification technology based on CSI.Principle Component Analysis(PCA)and Butterworth low-pass filter are employed to reduce dimension and denoise.Then the short time energy is used to detect human.Finally,RF algorithm is employed to classify nine features.The results shows that the identification accuracy can achieve 0.98 of True Positive Rate(TPR)from two persons.Even when there are a group of 10 in train set,the system can reach 0.92 of TPR.Those accuracies prove the effectiveness of identification technology based on Wi-Fi.(2)Another function of the identification technology is to achieve personal identity legal verification.Because the illegal human usually are not in train set,this is a big challenge to common classifiers.Considering the interference of illegal people to the identification system,the paper advances identity legal verification technology based on CSI to deal with the challenge.The discrete wavelet transform is employed to extract the approximate coefficient as features from the principal components.Then the K-means algorithm is used to cluster the known legal personnel data and get the clustering centers and distance thresholds.Finally,the distance between the threshold and the unknown person data to the cluster center is compared to verify identity legitimacy.Through evaluating CSI data from 10 volunteers in a real lab scenario which showed that the system can identify a person with TPR and True Negative Rate(TNR)of 0.98 to 0.80 from a group of 1 to 5 people respectively,it proved the validity of identity legal authentication technology based on CSI.