| Indoor positioning is the only way to realize the interconnection of all things.As a key feature of the information industry,indoor positioning technology presents a broad market prospect with the rapid technological breakthrough.However,limited by the impacts of complex environment,the current indoor positioning technology is in full swing but no consensus.With the widespread deployment of commercial wireless devices,the fine-grained channel state information(CSI)has received widespread attention with high development potential in the field of indoor sensing and positioning.In this thesis,a practical indoor positioning system based on enhanced CSI feature with perceptual adaptability(C-InP)is proposed to solve the three major problems of low resolution of CSI amplitude and phase characterization,the impact of dynamic human interference on positioning with the problem of high false alarm rate caused by non-line-of-sight propagation,and the contradiction between high precision requirement and system complexity.The main research contents and achievements of this thesis include the following three aspects:(1)Optimization and enhancement of CSI amplitude and phase characteristics.Aiming at the problem of low CSI resolution caused by noise interference of indoor environment and hardware defects on CSI measurements,we consider the abnormal filtering of CSI amplitudes in indoor scenes,realizing the amplitude outlier filtering algorithm of median variance probability(AMVP).Combined with calibration enhancement processing method of phase characteristics,we also implement the phase optimized calibration algorithm based on the least squares linear transformation(PLSC).The mapping relationship between the optimized high-resolution features and the corresponding locations is established.The enhanced CSI feature lays a signal foundation for the design of C-InP.(2)Lightweight dynamic sensing module design.In response to the real-time recognition requirements,the fuzzy feature granulation(FFG)method is proposed from time dimension to complete the research goal of lightweight detection operations.Moreover,aiming at the problem of high false alarm rate caused by the influence of NLOS propagation on the detection accuracy of moving human,B-SVC method is utilized to realize the high-precision recognition.In respect of the influence of signal fluctuation caused by the dynamic human interference on positioning,this thesis proposes a lightweight dynamic sensing module to identify human interference which is prior to the positioning module in system.(3)The indoor positioning system based on enhanced CSI feature with sensing adaptability.With regard to the contradiction between the high-precision requirements and the low-complexity of the positioning system,we propose a practical indoor sensing and positioning system,which consists of two parts:lightweight dynamic sensing module and positioning module.The positioning module weighs the accuracy and complexity,adopts SVD to optimize the dimension of CSI feature matrices,and then utilizes M-SVC method to realize indoor positioning.Finally,from the above three aspects of the characterization ability of the enhanced CSI features,the performance of sensing and positioning module,and the accuracy and complexity of C-InP,comprehensive experiments and comparative verifications are carried out in four typical indoor scenes to evaluate the advantages of C-InP.The experimental results show that the AMVP and PLSC algorithms proposed in this thesis are far superior to the traditional processing methods,and the performance of C-InP is better than the existing MDS-KNN and NB systems.The detection accuracy of integrated environment and complex garage is 89.4%and 94.7%,respectively,and the corresponding average distance error is respectively 0.49m and 0.81m.In terms of the average running time of the positioning system,C-InP achieves a performance improvement of about 23.2%compared with the single positioning module. |
PDF Full Text Request