Indoor Human Detection Base On Channel State Information

With the network technology broad application,wireless communication technology has been rapidly developed,WLAN has become very popular,schools,shopping malls,coffee shops,households can be connected to WIFI signals,these widely installed WIFI devices provide a physical basis for WIFI-based situational awareness technology.Radio waves can not only transmit data,but also perceive the environment.Traditional environment sensing technology based on WIFI signal is usually studied based on RSSI,but it can not provide fine-grained information because it is limited by multipath effect.In recent years,with the availability of channel state information(CSI)for commercial devices,many studies have begun to focus on CSI.At present,the application research based on WIFI mainly includes indoor personnel detection,indoor positioning,motion recognition and so on.In this thesis,we focus on environmental sensing technology,then several commom environment sensing techniques are summarized,and introduces the technology of personnel detection,indoor positioning and motion recognition based on WIFI.Based on this,this thesis proposes a CSI-based indoor personnel detection scheme,which includes two main functions: indoor human identification and number detection.In the CSI-based indoor human identification scheme,firstly,the CSI signals of the tester in both moving and stationary states are collected,then the original CSI signals are denoised by low-pass filtering,and the dynamic and static features are extracted respectively.The static features use the time-domain mean,the dynamic features use the frequency-domain kurtosis,and the support vector machine uses the kernel function as the radial basis.This two-dimensional feature is classified to distinguish different people.Experiments are carried out under different scenarios and conditions.The experimental results show that the accuracy of target recognition can reach 93% in 10-person system tests.In the CSI-based indoor number detection scheme,we use two antennas to receive signals.Firstly,the signals of the tested area are collected,then the original CSI signals are denoised by low-pass filtering and PCA.Then,the processed data are intercepted by sliding windows,and the data are divided into stable data sets and fluctuating data sets by calculating the correlation of adjacent windows.Then the mean and standard deviation of each data set are extracted as the feature of each windows.Then the two-dimensional features are classified by support vector machine.The data set is divided into two parts: stable and fluctuating.Then the ratio of fluctuating data and actual number is 21:1 by analyzing the relationship between fluctuating data and actual number of people.We validate the system under different conditions,and the results show that the accuracy of the number detection scheme of the system can reach more than 95%.