Research On Wireless Indoor Location Method Based On RSS And CSI Feature Fusion

An important feature of the mobile Internet and the Internet of Things is the ability to provide location-based services.However,GPS signals cannot penetrate most building materials and cannot be used for indoor positioning effectively.WiFi signals exist in indoor spaces widely and are an ideal positioning source,the fingerprint positioning based on the WiFi signal has higher accuracy and a more accurate positioning result can be obtained than a single fingerprint by fusing different fingerprints.This thesis focuses on the issues that Received Signal Strength and Channel State Information fingerprints cannot be combined effectively and there are a lot of meaningless fingerprint matching calculations during the positioning process.The main research contents are as follows:The accuracy of the CSI fingerprint affects the accuracy of the positioning system greatly.Therefore,this thesis first studies the preprocessing method of the CSI data.Transforming the CSI data into the time domain,by eliminating the low power path,the purpose of reducing interference in the CSI signal is achieved.According to the principle of coherent bandwidth,the 30 subcarriers of the CSI signal are equivalent to 4 independent subcarriers,and the CSI amplitude is averaged on each independent subcarrier to reduce the dimensionality of the CSI signal while retaining the characteristics of the CSI information.Then,by introducing a distributed CSI fingerprint generation algorithm,a CSI fingerprint database with low dimensions and high accuracy is generated.Proper fusion of RSS and CSI fingerprints feature can get more accurate positioning results.Therefore,the confidence of each candidate reference position is defined in this thesis to reflect the sparseness of the candidate reference point distribution generated by RSS and CSI fingerprint positioning algorithms.By modeling the similarity between the RSS and CSI fingerprint positioning results,the value of confidence is generated to achieve the fusion of RSS and CSI fingerprints.Based on this,an RSS and CSI Fingerprint Fusion algorithm is proposed.Simulation results show that the fingerprint fusion algorithm has higher positioning accuracy than the existing algorithms.For the issue of a large number of meaningless fingerprint matching calculations in the localization process,in this thesis,the fingerprint database is clustered and the clustering results are used for coarse positioning,which reduces the number of fingerprint matching calculations greatly.However,the combination of the existing clustering algorithm and the positioning system will cause a certain degree of positioning accuracy loss.Therefore,this thesis defines a fingerprint based on CSI signal multipath delay and uses it as the external information of the clustering algorithm.The multipath delay characteristic contained in the fingerprint reflects the physical spatial information of different receiving locations,so the accuracy of clustering can be improved.Combining the Affinity Propagation clustering algorithm with RCF,this thesis researchs an AP Clustering Fusion Positioning Algorithm.By simulating a large number of measured data,it is verified that the algorithm further improves the positioning accuracy under the premise of lower system complexity.