Research On Indoor Device-free Passive Localization And Its Applications On Behavior Monitoring

With the rapid growth in location service for the indoor environment,the indoor positioning based on fingerprint recognition has attracted wide attention because of its high accuracy.The received signal strength indicator(RSSI)is widely used as a navigation system and positioning system as a conventional scheme,but the multipath effect caused by complex indoor environment leads to a reduction in the accuracy of the system.In recent years,the channel state information of a physical layer(CSI)can be obtained by more commercial wireless devices,which can show a more granular characteristics of the signal,and has better stability.In this thesis,an indoor device-free passive positioning algorithm based on CSI fingerprint is proposed,which can estimate the specific coordinate position of the target more accurately.Firstly,DBSCAN is used to remove the noise in the original data and reduce the interference of the outliers.And then we use the principal component analysis(PCA)to extract the project with high contribution rate in the feature which can reduce the feature dimension and computational complexity.Finally,the association model of CSI fingerprint and position coordinates is established by using the regression algorithm of support vector machine(SVM).In addition,CSI fingerprints are also applied to behavior monitoring.In intrusion detection,we used the SVM binary classification method for detecting intrusions.In simple target recognition,multi-classification method of SVM is used to distinguishing the target.In the people counting,we use the weight-based dilated matrix method combined with SVM regression algorithm to calculate the number of targets.In the population density detection,the dynamic time normalization(DTW)algorithm is used to match the optimal population density.The experimental results show that the mean error localization distance of the localization algorithm proposed in this thesis is 1.37 meters,and it is proved that the method has obvious advantages in positioning accuracy by comparison with various positioning methods.In the intrusion detection,the door intrusion detection and the room detection has the accuracy of 98.2% and 99.1% respectively.In the simple target recognition,the classification accuracy rate is 98.7%.The mean error of people counting is 0.62.The accuracy rate of population density is 95%.Experiments show that the behavior monitoring based on CSI fingerprints is effective and high accuracy.