Study On Methods Of Parameter Estimation Of Polarized Array Signal In Alpha Stable Noise

Polarized array signal processing is a new relative research direction in the field of signal processing. In the last decades, the research based on DOA and polarization estimations of array signals has became increasely important in the area of signal processing, which has a broad application prospects on Radar, communications, hydro-meteorological, geological exploration, pattern recognition and so on. In this thesis, regarding the electromagnetic vector sensors as the research subject, the issues of DOA and Polarization estimations of array signals, which are based on the background of theαstable distribution noise, are researched.Most of the traditional algorithms of the electromagnetic vector sensor signal processing are based on Gaussian noise, which is added as a common background noise signal, such as MUSIC and ESPRIT, etc. But these algorithms are ineffective for non-Gaussian noise. In recent years, we found a special kind of noise which has characteristics of very large peak pulse in the electronic warfare and communications systems. This kind of noise can be described by theαstable distribution. Therefore, it is necessary to research electromagnetic vector sensor signal DOA and polarization estimation based on the background ofαnoise, which is also the focus of this study.αStable distribution is only a class of distributed which satisfy the generalized central limit theorem. When the timeα=2,αstable distribution is changed into a Gaussian distribution. Obviously, Gaussian distribution is a special case ofαstable distribution. Noise in nature can be described effectively byαstable distribution. However, the biggest difficulty to researchαstable distribution is that it does not exist more than second-order moment and higher order moments. That is, it does not have finite variance. This makes a lot of algorithms, which were used to deal with the Gaussian noise applied to the background ofαstable distribution noise effectively, appear degenerate or even the wrong result. In this paper, the fractional low order loop correlation function based on the attenuation facto is proposed to research the issues of DOA and Polarization estimations of array signals, which are based on the background of theαstable distribution noise. The research of this thesis is based on theαstable noise. And the main studies are about estimations of DOA and polarization of electromagnetic vector sensor signal, linear polarization array signal and arbitrary structure of polarization array signal. Firstly, the signal model of electromagnetic vector sensors is introduced. Secondly,αstable noise has p-order moment. Fractional low order circular correlation can separate signals of various frequency, and suppress stable noise signal of non-cyclostationary. Finally, the singular value of the fractional low order cyclic-related function is decomposed, and the signal space and the noise vector space are obtained. In the meantime, the DOA and polarization estimation of polarized array are estimated in the noise vector space. The simulation results shows that the fractional low order cyclic correlation-based method can suppress theαstable noise, and improve signal parameters of performance effectively.