Research On Fingerprint Optimization Method In Indoor Localization

With the rapid growth of people's dependence on indoor localization services,WiFi fingerprint localization technology has a good application prospect with the advantages of easy deployment and low cost.For fingerprint localization method,fingerprint and localization algorithm directly determine the accuracy of localization.However,in the practical localization process,there are some problems,such as low fingerprint quality and poor fingerprint interpretation ability.On the one hand,the traditional received signal strength(RSS)fingerprint localization method must ensure that the RSS fingerprints in the offline database and the online RSS fingerprints are independently and identically distributed.However,the RSS fingerprints are often affected by factors such as environmental changes and heterogeneous devices,resulting in the decline of fingerprint quality.On the other hand,the traditional channel state information(CSI)fingerprint localization method has the problems of single fingerprint feature and incomplete hardware error optimization,resulting in poor fingerprint interpretation ability and low fingerprint quality.To solve the above problems,this thesis studies the fingerprint optimization method in indoor localization.The specific research work is as follows:This thesis proposes a subspace domain adaptation fingerprint optimization indoor localization algorithm,which improves the localization performance by optimizing the fingerprint quality.In order to reduce the distribution difference between the source domain and the target domain,fingerprint reconstruction is introduced.By learning a projection matrix,the source domain and the target domain are transfer to a common subspace.In the common subspace,each target domain fingerprint can be represented by the source domain fingerprint with the reconstruction coefficient matrix;In order to improve the discrimination in the common subspace,the label distribution alignment is constructed based on the source domain label and the target domain pseudo label;When only the fingerprint reconstruction and label distribution alignment are used to optimize the fingerprint,the inherent geometric structure of the fingerprint may be destroyed.In order to preserve the local geometric structure relationship between fingerprints,that is,adjacent fingerprint samples often have the same label,the local geometric structure constraint is introduced to promote the stable fingerprint optimization.The experimental results verify the effectiveness of the proposed algorithm.Aiming at the shortcomings of traditional CSI fingerprint localization methods:(1)the dimension loss of single frequency band or single domain CSI is large and the fingerprint interpretation ability is poor;(2)the baseband design of hardware equipment leads to CSI amplitude and phase distortion and poor fingerprint quality;(3)a single fingerprint sample may lead to poor location robustness,this thesis proposes an indoor localization algorithm based on dual-band WiFi time-frequency domain j oint fingerprint optimization.Firstly,the amplitude and phase of dual-band CSI are optimized,and then the j oint fingerprint of dual-band time-frequency domain is extracted.The joint fingerprints of multiple samples are input into the localization model to construct the location candidate set.Then,according to the candidate set,a trustworthy location selection algorithm is proposed to jointly optimize the kernel density function and weight of each candidate location to obtain the optimal estimation of the final location.Experimental results in two real-world environments verify the effectiveness of the proposed algorithm.