New Algorithms For Emitter Localization Using Multiple Receivers

Emitter localization with multiple receivers means the unknown location of the emitter is estimated from the received signals in receivers. Emitter localization is widely applied in electrical reconnaissance, environmental monitoring and other areas. Such as passive source localization and wireless sensors network localization. The algorithms in this dissertation contain two types. The first one is the common two steps localization method where the emiter location is estiamted by some kinds of measurements. Such as time of arrival(TOA), Angle of arrival(AOA), time difference of arrival(TDOA) and frequency difference of arrival(FDOA). These measurements are extracted from the received signals. The other type is the direct position determination(DPD) method. The DPD method determinates the emitter location from the received signal directly.Chapters 2-4 are about the two steps localization. Chapter 2 investigates source localization using FDOA only. Chapter 3 is about sources localization and calibration using TDOA and FDOA. Chapter 4 considers relative source localization with TOA. Chapter 5 focuses on the direct position determination method based on delay and Doppler. The main content of this dissertation are detailed as follows.In chapter 2, source localization using FDOA measurements only is investigated. Two different localization geometries are studied. First, we focus on the localization geometry where source location is constrained by the earth surface. We first derive the Cramer-Rao lower bound(CRLB) for source location estimation. Then we propose a linearizing method which convert the localization model to a linear least squares estimator with a nonlinear constrained. The Gauss-Newton iteration method is developed to conquer the source localization problem. From the analysis of solving Lagrange multiplier, the algorithm is a generalization of linear-correction least-square estimation procedure under the condition of geolocation using FDOA measurements only. In the second localization geometry, there is no sphere constraint. We focus on effect of sensor location error. We first perform a CRLB and mean-square error(MSE) analysis for the source localization. An iterative algorithm is applied to estimate the locations of the source and sensors. Simulation results show that the estimation accuracy approximately attains the CRLB and demonstrate the feasibility of the accuracy analysis.In chapter 3, sources localization and calibration using TDOA and FDOA measurements are investigated. There are three different localization geometries. In the first geometry, there are one source and one calibration emitter. The TDOA and FDOA measurements from the calibration emitter are used to modify sensor locations. In the second geometry, there are many sources. Some sources have prior location information, but the prior information is also inaccurate. The TDOA and FDOA measurements from the sources with prior information are used to modify sensor locations. In the last geometry, there is only one unknown source and no calibration emitters. But some sensors are calibration sensors. The TDOA and FDOA measurements from the calibration sensors are used to modify sensor locations. In each localization geometry, the modified sensor locations and the TDOA/FDOA measurements from the source are used to estimate the unknown source location. Simulation results show the proposed algorithms can attain their CRLBs respectively when TDOA and FDOA measurements noise is small.In chapter 4, relative sources localization is considered. In this geometry, the locations of all sources and sensors are unknown. It is unavailable to estimate the locations of the sources and sensors using TOA measurements only. However, the relative locations of the sources and sensors can be estimated. We first derive the CRLB of the relative source localization. Then bilinear method is used to convert the nonlinear equations to linear ones. Singular value decomposition(SVD) is used to transform the unknown parameters of locations into elements in a mixing matrix. This technique reduces the dimension of unknown parameters significantly. Finally, least square estimator is used to estimate the elements in the mixing matrix. Moreover, we also derive the weighting matrix for the least square estimator. So the localization accuracy can be improved by the weighted least squares estimator.In chapter 5, coherent summation DPD method is investigated. Intercepting multiple signal intervals in each receiver has been a common technique to improve source localization accuracy. However, the fusion of those signal intervals in each receiver has not been investigated successfully. In this chapter, we propose a new coherent summation approach for DPD based on time delay and Doppler shift. The method takes into account phase differences among multiple signal intervals captured by each receiver. Both theoretic analysis and computer simulation demonstrate that coherent summation technique can improve localization accuracy significantly over previous normal summation method. In addition, due to the availability of the signal waveform in receivers, we also compare the performance differences between unknown waveform and known waveform, which are corresponding to passive and active localization respectively.