| Wi-Fi-based indoor sensing technology has received widespread attention due to the advantages brought by the widespread deployment of Wi-Fi devices.With the improvement of Wi-Fi IEEE 802.11n protocol,the more granular Channel State Information(CSI)at the physical layer has replaced Received Signal Strength Indication(RSSI)serves as the basis of Wi-Fi sensing technology.However,there are also many problems like relatively low temporal and spatial resolution caused by various phase errors and spectral discontinuities.for CSI-based Wi-Fi sensing technology.Although the use of machine learning and deep learning algorithms can enrich the application diversity of Wi-Fi sensing,the accuracy of Wi-Fi sensing is not improved from the bottom layer.This thesis mainly focuses on the defect of CSI-based Wi-Fi sensing technology with low sensing accuracy because of insufficient bandwidth and research on an ultra-wideband spectrum splicing technology that can increase Wi-Fi bandwidth for sensing.The main work of this thesis is as follows:1.Analysis and elimination of CSI linearity error.Wi-Fi signals are susceptible to noise during transmission,which will cause errors in CSI data.Therefore,eliminating errors in CSI data has become a prerequisite for improving sensing accuracy and spectrum splicing technology.Based on the CSI phase model,this thesis analyzes the linearity error in the CSI such as packet detection delay,sampling frequency offset and carrier frequcncy offset,and proposes corresponding elimination algorithm.2.Source research and elimination of CSI non-linearity error.In addition to linearity error,non-linearity error is also found in experiment in this thesis.In order to research its source,this thesis used QCA 9300 as a representative of commercial network card.First,this thesis restored the baseband architecture design of QCA 9300 based on public data,and made reasonable guesses the source of non-linearity error according to the design,then through a large number of experiments,it is found that source of non-linearity error is related to the baseband of the network card,especially the baseband filter.At last this thesis proposes an algorithm based on the least squares algorithm to eliminate it.3.High-precision ultra-wideband spectrum splicing algorithm.After eliminating all types of error,this thesis researches an ultra-wideband spectrum splicing algorithm based on highprecision CSI data.Splicing the original continuous narrow spectrum measurement based on the subcarrier phase consistency constraint,then splicing the wide spectrum again based on the time domain response consistency.Finally eliminates the overall splicing error to obtain an ultra-wideband spectrum.Finally,the CSI data is collected on a commercial network card platform based on the 802.11n protocol for experiments,the error elimination algorithm and the spectrum splicing algorithm are used to process the collected CSI,and the spliced CSI is compared with the original CSI.The experimental results show that the bandwidth of the spliced CSI is much wider than the original CSI,and the phase spectrum becomes smooth.When using the spliced CSI for ranging,the result also shows that the accuracy and resolution are improved. |
