Study On Channel Modeling And Signal Parameter Estimation For HF Communication Based On Alpha Stable Distribution

Reasonable description and handling of noise and interference, by way of reducing its adverse effects and improving algorithm performance, has always being a hot field of research in signal processing. With gradually in-depth understanding of noise, it was found that in many real environments, noise and interference are not completely in line with the Gaussian distribution, but has trailing pulse characteristics of alpha stable distribution, which makes the conventional Gaussian-hypothesis-based algorithms no longer suitable for use, leading to remarkable performance degradation. In HF communication, as the ionospheric channels are vulnerable to outbreaks of sunspots and atmospheric thunderstorms, the frequency of the channel dispersion as well as the channel noise is also of trailing characteristics. Thus channel modeling and the corresponding new signal processing approaches for impulse noise in HF environments are of important theoretical significance and practical value.In this thesis, with the hypothesis of alpha stable distribution of noise, a series of studies is made including the improvement of the currently existing HF ITS channel model, the time delay estimation (TDE) of HF fading signal, the estimation of direction of arrival (DOA) and the symbol rate estimation for MPSK signals. Furthermore, some new design ideas and statistical signal processing methods are put forward. The innovative achievements obtained in this thesis are summarized as follows.(1) In respect of the modeling and simulation of HF ionospheric channel, an improved model of HF broadband channel is proposed based on the assumption of alpha stable distribution. Targeting at the problems of less operational flexibility and higher computational complexity of existing ITS HF broadband models, and by the use of above proposed model, many improvements are made both in generation of random modulation function of ITS model and the additive noise model. A new formation method based on the symmetric alpha stable distribution (SαS) function transform was proposed for generating random modulation function, which can not only successfully model the narrowband noise power spectrum shaping but also extend the flexibility of Doppler Shift, hence is more suitable for simulations of various typical HF channel transmissions; The new SαS-based additive noise model simplifies the noise generating procedure and shows a bettered conformity with actual situations; An improved approximation is obtained for time delay power distribution based on the anchor point fitting approaches. Simulation results for comparison between the scattering function diagram and the typical transmission channel show that the proposed model is more flexible and accurate in describing the major statistical properties of HF ionospheric channels.(2) As regards time delay estimation for HF fading signal in the impulsive noise environment, two improved time delay estimation algorithms are proposed against the influence of signal fading that was not considered in the original double-base mode of time delay estimation. Firstly, based on the fact that the nonlinear transform can restrain abrupt peak changes of impulsive noise, a TDE algorithm based on the combination of nonlinear transform and out-of-band noise filtering was proposed for time delay estimation of HF fading signals in impulsive noise environments and outperform the conventional time delay estimation algorithms based on both the fractional lower order statistics (FLOS) and the relative entropy criterion (GCA). Secondly, A concept of Nth order nonlinear compressed core function (NCCF) along with a corresponding NCCF-based TDE algorithm was proposed. And a verification of the fact was provided that the statistics based on this function have the similar boundary characteristics to that of the fractional lower order statistics. Simulation results have proved the feasibility and the improved estimation performances of the algorithm over the conventional FLOS TDE, GCA TDE algorithm, etc.(3) In regard to DOA estimation in impulsive noise environment, as we know, because SaS does not possess second order statistics, DOA estimation algorithms based on MUSIC (multiple signal classification) could hardly obtain proper eigenvalues in covariance matrix decomposition, leading to performances degradation of DOA estimation. Aiming at this background, two improved algorithms of DOA estimation were proposed. They are MUSIC algorithm based on normalized compression function and MUSIC algorithm based on nonlinear compressed core function (NCCF). Both of them adopt the compression method to restrain the peak tailing of impulsive noise, and to form a bounded matrix similar to the covariance matrix. The corresponding signal subspace and noise subspace are spanned through the bounded matrix eigenvalue decomposition. And then, a peak search of space spectrum is implemented using the orthogonality of the subspaces for DOA estimation. Simulation results have validated the feasibility and performances stability of the above two DOA estimation algorithms.(4) In order to improve the accuracy of symbol rate estimation for MPSK signals, a new symbol rate estimation algorithm in alpha stable noise environment is proposed. By using normalized compression function to compress the observed signals, the impulse noise influence on the cyclic spectral density is restrained effectively. By means of the maximal value of cyclic spectrum density at the section of f= 0, the search scope of the cyclic frequency sectional is confirmed. The accumulated amount of the cyclic spectral density in the scope results in the estimation of the symbol rate. Simulation experimental results show the new algorithm can achieve better stability and accuracy than the existing algorithms of symbol rate estimation.