Indoor Localization Research Based On GBSA-MDL Source Number Estimation And 2D Multi-Sensor Spatial Spectrum Fusion

Direction-of-Arrival(DOA)or Angle-of-Arrival(AOA)estimation of multiple sources using an array is a fundamental problem in modern signal processing.However,compared with different DOA estimation methods,the subspace-based methods are the most commonly used due to their high resolution and simplicity of computational.Meanwhile,in order to separate between source subspace and noise subspace which are the compositions of the signal impinging the array elements on the receiver side,subspace-based algorithms require the exact information of the source number which is usually unknown.Accordingly,several classical source number estimation methods have been proposed in the past based on information theoretic criteria such as Minimum Description Length(MDL).But,in most known real applications there is a scenario in which the number of sensors goes to infinity at the same speed as the number of snapshots(general asymptotic case)which yields to a blind performance for the classical MDL and results in an inaccurate source number estimation.In the other side,there is a well-known metaheuristic optimization algorithm which mimics the spiral arm of the spiral galaxy in the outer space in order to find a precise solution to hardly optimized problems named as Galaxy Based Search Algorithm(GBSA).In the first part of this thesis,the GBSA algorithm is modified and applied with the MDL criteria so as to optimize and correct the diction of source number under the general asymptotic case.Several simulation results show that the proposed GBSA-MDL algorithm gives reliable results compared to several used source number estimation methods.In the second part of the thesis,a novel 2-D Multi-Sensor Spatial Spectrum Fusion(2DSSF)localization algorithm based on the Multiple Signal Classification(MUSIC)method is proposed.We know that in a starving indoor environment,the propagation of the source signal is strongly attenuated by reflection when it hits the surface of an obstacle,which results in the high existence of Non-Line of Sight(NLOS)signals arriving at the receiver through different paths then,localization using traditional spatial spectrum estimation techniques easily fails due to low Signal to Noise Ratio(SNR).So,in the 2D-SSF algorithm,the output data of each Uniform Rectangular Array(URA)at each Access Point(AP)are first processed to get the noise subspace data.Then,after estimating the corresponding azimuth and elevation angles of each array using the MUSIC approach and finding the position of each point relative to each sensor in the search grid with the help of grid refinement algorithm,the parameters of interest of the target are estimated from a single spectrum that results from fusing all maximum noise subspaces where the position corresponding to the minimum error between the set of angles and every estimated point in the searching area is situated.Different simulation results of the proposed method in terms of RMSE as a function of SNR for various APs LOS/NLOS scenarios,the change in the number of antennas at each AP and the comparison with the MUSIC approach and the 1-D localization based spatial spectrum fusion algorithm are carried out.The obtained results proved the significant performance of the proposed 2D-SSF localization algorithm with the strong presence of NLOS signals.