Classification and localization of electromagnetic and ultrasonic pulsed emitters

[pdf] Classification and localization of electromagnetic and ultrasonic pulsed emitters
Outstanding Thesis Award Recipient 2017-2018 Edition
Director(es): Robles, Guillermo
Departamento/Instituto: Universidad Carlos III de Madrid. Departamento de Ingeniería Eléctrica
Titulación: Programa Oficial de Doctorado en Ingeniería Eléctrica, Electrónica y Automática
Fecha de edición: 2017-12
Fecha de defensa: 2017-12-18
Tribunal: Presidente: Andrea Cavallini.- Secretario: José Antonio García Souto.- Vocal: Iliana Portugués Peters
Palabras clave: Ultrasonic pulsed emitters , Electromagnetic waves , Radiative sources , Signal processing , Algorithms , Monte Carlo methods
Derechos: Atribución-NoComercial-SinDerivadas 3.0 España

Abstract  – The localization of radiative sources is very important in many fields of work such as: sonar, radar and underwater radar, indoor localization in wireless networks, earthquake epicenter localization, defective assets localization in electrical facilities and so forth. In the process of locating radiative sources exist many issues which can provoke errors in the localization. The signals acquired may belong to different sources or they can be mixed with environmental noise, then, their separation before using localization algorithms is of great interest to be efficient and accurate in the computational process. Furthermore, the geometry and radiation characteristics of the receivers, the nature of the signal or their measuring process may cause deviations in the signal onset calculus and therefore the source localization could be displaced from the actual position. In this thesis, there are three kinds of algorithms to undertake three steps in the emitter localization: signal separation, onset and time delay estimation of the signals and source localization. For each step, in order to reduce the error in the localization, several algorithms are analyzed and compared in each application, to choose the most reliable. As the first step, to separate different kinds of signals is of interest to facilitate further processing. In this thesis, different optimization techniques are presented over the power ratio (PR) maps method. The PR uses a selective spectral signal characterization to extract the features of the analyzed signals. The technique identifies automatically the most representative frequency bands which report a great separation of the different kinds of signals in the PR map. After separating and selecting the signals, it is of interest to compare the algorithms to calculate the onset and time delay of the pulsed signals to know their performance because the time variables are inputs to the most common triangulation algorithms to locate radiative and ultrasonic sources. An overview of the algorithms used to estimate the time of flight (ToF) and time differences of arrival (TDoA) of pulsed signals is done in this thesis. In the comparison, there is also a new algorithm based on statics of high order, which is proposed in this thesis. The survey of their performance is done applied to muscle deep estimation, localization in one dimension (1D), and for the localization of emitters in three dimensions (3D). The results show how the presented algorithm yields great results. As the last step in the radiative source localization, the formulation and principle of work of both iterative and non-iterative triangulation algorithms are presented. A new algorithm is presented as a combination of two already existing improving their performance when working alone. All the algorithms, the proposed and the previous which already exist, are compared in terms of accuracy and computational time. The proposed algorithm reports good results in terms of accuracy and it is one of the fastest in computational time. Once the localization is achieved, it is of great interest to understand how the errors in the determination of the onset of the signals are propagated in the emitter localization. The triangulation algorithms estimate the radiative source position using time variables as inputs: ToF, TDoA or pseudo time of flight (pToF) and the receiver positions. The propagation of the errors in the time variables to the radiative source localization is done in two dimensions (2D) and 3D. New spherical diagrams have been created to represent the directions where the localization is more or less sensible to the errors. This study and their sphere diagrams are presented for several antenna layouts. Finally, how the errors in the positioning of the receivers are propagated to the emitter localization is analyzed. In this study, the effect in the propagation of both the relative distance from the receivers to the emitter and the direction between them has been characterized. The propagation of the error considering the direction is also represented in spherical diagrams. For a preferred direction identified in the spheres, the propagated error in the source localization has been quantified regarding both the source distance and the magnitude of the errors in the receivers positioning.