Research On High Resolution Indoor Positioning Method Based On Spatial Spectrum Fingerprints

Recently, wireless communication technology has attracted vast attention for the growing demand of indoor location. Various environmental factors such as moving people and multipath effect degrade the positioning performance in the complicate indoor environment, thus the accuracy and robustness of current indoor positioning algorithm could not meet the people’s requirements. The available indoor positioning systems, are restricted to single function, poor expansibility, or are difficult for widely using with high price. Therefore, most scholars are looking forward to some indoor positioning systems with low cost and sufficient detection of environmental information.In this thesis, the algorithms proposed are mainly based on the way of information fusion operating on the spatial spectrum fingerprints and verified through the data collected from USRP hardware platform. It is proved that the algorithms proposed can achieve a satisfying result.The main contents of this paper are as follows:Firstly, the thesis introduces some classic indoor positioning theories such as TOA, TDOA, RSS and DOA, and the application of these theories in linear and nonlinear algorithm. Then describes the advantages and disadvantages of current frequently used indoor positioning systems. With the analysis of current indoor positioning algorithms and systems, we have gained a useful insight into the disadvantages of current indoor positioning technology, and it provides the improving direction for us.Secondly, for the shortcomings of the RSS fingerprinting used in the indoor environment, this paper puts forward the principal components analysis(PCA), covariance fingerprint and the four order cumulant fingerprint. Compared with the RSS fingerprint, the fingerprints save more indoor space information, thus the robustness is better, and it is verified through simulation experiment.Thirdly, this paper introduces a data clustering technique, which reduces the calculation amount by the data clustering matching in fingerprint database. In order to make full use of a variety of fingerprints in the indoor environment, this paper introduces the Ensemble learning, neural network and DS theory to carry on the integration of various kinds of fingerprints, and verify its performance through simulation experiments.Finally, this paper designs a set of indoor positioning system based on the open source software radio platform. The system can sample the complete information of the interior space. Then, the system collects information in the indoor environment, and use the data to verify the proposed algorithm.