Research Of Near Field Source Localization Method Based On Higher Order Statistics And Fractional Lower Order Statistics

The source localization is one of the most important research aspects in arraysignal processing. Based on the sources in near field or far field of the array, thesource localization can be divided into two parts. It includes direction of arrivalestimation of far field and near field source localization. The DOA estimation of farfield has abundant applications in many engineering domain, such as mobilecommunication, radar, sonar, etc. In the past thirty years, the scholars of the world haddone a lot of research on DOA estimation of far field and achieved plentiful fruit. Ingeneral, the application domains of DOA estimation of far field have also existedproblems of near field sources localization. For example, the mobile communicationlocalization of short distance, sonar, voice sources localization in room and so on.Because the sources of these statuses are in near field region of the received array,they belong to near field sources localization.Unfortunately, the researches about near field source localization have not beensufficient in academe. When the near field source is in complicated noise environment,the traditional localization algorithms have low estimated precision. It is also nosufficient for the other parameters estimation research, such as carrier wave frequencyis unknown and three dimensional space near-field sources. Therefore, the researchcontents of our paper are developed based on these problems.The innovation points of this dissertation focus on the three aspects as follows.Firstly, in order to restrain impulsive noise, the near field 2-D rooting MUSIC methodbased on fractional lower order statistic is proposed. The experiment results show thatthe new method can restrain impulsive noise's influence and estimate azimuth(bearing) and range parameters of the near field sources in effect. Secondly, byexploiting agility of definition and blind Gaussian property of the higher ordercumulant, the near field 2-D polarization Unitary ESPRIT method, the jointlyestimation method of azimuth, range and frequency based on four-order cumulant andthe 3-D like ESPRIT method are proposed which expanded the aperture of the array.These proposed methods have good estimation precision of parameter. Finally, basedon polarization information of the polarization sensors or time-space information, thenear field TLS-ESPRIT method by exploiting the polarization sensors, the algorithmsof jointly estimation azimuth, range and frequency based on second order statisticsand 3-D time-space algorithms are addressed, which have good performance andsmall compute burden for using second order statistics. So it solves the contradictionbetween estimate performance and compute burden.The dissertation consists of eight chapters.Chapter one is introduction, in which the main methods of near field sourcelocalization are analyzed and summarized. Then the parameters estimation methods ofthe two-dimensional near field source (azimuth and range), the three-dimensional nearfield source (azimuth, elevation and range) and the jointly estimation of frequency,azimuth and range are discussed respectively. The meaning of the theory and thepractical value is elucidated.Chapter two is the fundamental knowledge about this dissertation. Theknowledge of matrix algebra is introduced. The definitions and properties offractional lower order statistics and higher order statistics are described.Chapter three discusses the problem of the near field source localization inimpulsive noise. The near field 2-D rooting MUSIC method is proposed. Due to usefractional lower order statistics to restrain impulsive noise, the proposed method has asatisfying result in impulsive noise environment.The near field source localization problem based on dual-polarization sensors isintroduced in chapter four. In order to improve estimation accuracy of the traditionalalgorithm, dual-polarization sensor is used as sensors of the received array. It canimprove estimation performance by using polarization information of polarizationarray. The near field source localization problems are discussed on two aspects byusing different mathematics tools, namely correlation function and four-ordercumulant. One method based on correlation function has the low compute burden andcovariance. The proposed method is adapted to additive Gaussian white noiseenvironment. However the proposed method based on four-order cumulant may be abetter choice in Gaussian color noise environment. The researches that the near fieldlocalization methods based on dual-polarization sensors have excellent accuracy andstabilization.Chapter five discusses the problem of the jointly estimation frequency, azimuthand range. In chapter three and chapter four, the researches of near field sourcelocalization are under the assumption that the carrier wave frequency of the source isknown. In this chapter, when the frequency of the source is unknown, the parameterestimation problem of near field source is described. In this chapter, a simple methodis proposed to estimate three-dimensional parameters, which has a small computeburden. For the sake of the accuracy of estimation, another algorithm is introduced byexploiting time domain and space domain information of the array. It is proved thatthe new methods can estimate three-dimensional parameters effectively.In chapter six, the problem of parameter estimation in three-dimensional spacedomain is discussed. From the point of view of the space domain, the discussedproblems in chapter three, chapter four and chapter five belong to two-dimensionalspace. In general, the sources lie in three-dimensional space. It is to say that near fieldsource localization method need to estimate azimuth, elevation and range of thesource. In order to reduce hardware burden, this chapter develops an effective methodto reduce sensors of the array by exploiting four-order cumulant. At the same time, atime-space localization method is proposed to improve estimation precession.In chapter seven, the time delay estimation problem of microphone array isconsidered. By analyzing the shortage of traditional method, a novel time-delayestimation method using multi-channel microphone array is proposed. Compared withthe traditional methods, the proposed method has much better performance on theabilities of anti-noise and anti-echo.In chapter eight, a brief summary of the dissertation is given, and the future workis also expected.