Study And Application Of Correntropy Based Time Delay Estimation Under Alpha-stable Noise

This thesis is mainly concerned with the time delay estimation (TDE) issue under the alpha-stable distribution noise environment by studying the conrrentropy theory. The research object is to find new TDE algorithms and improve traditional algorithms based on the conrrentropy theory. The TDE weighs very much in terms of the extraction of model parameters in the signal processing fields and it is of great meaning to the topic of TDE both in theory and application. As one of the promising ways of development in the TDE fields, the TDE under impulsive noise environment has been given great attention by people. Usually, the impulsive noise model is described by alpha-stable distribution and because of the direct ratio between minimum dispersion and fractional lower order moment, the theory of fractional lower order statistics has become the theoretical principle to solve the TDE problem under impulsive noise background. However, it is not the only way.Correntropy, as a measure of two random variables, is a kind of statistics moment which could also reflect the information of dispersion coefficient. In this thesis, the possibility to solve the TDE issue with correntropy has been discussed and several TDE methods have been proposed with details as follows:Firstly, the equivalency of the maximum correntropy criterion and the minimum dispersion criterion is derived from the parametric representation for zero location SaS distributions. This result is used to propose the adaptive time delay estimation under SaS noise background.Secondly, considering the fact that the self-adaptive algorithm based on conrrentropy could not give satisfying TDE results under impulsive noise background, two improved MCC methods has been proposed. Theoretical analysis and simulation experiments show that these two estimation methods can give satisfactory results.Thirdly, the traditional algorithms time delay estimation (TDE) underα-stable distribution noise conditions are usually based on the fractional lower order statistics (FLOS), This paper proposes a generalized correntropy function (GCF) algorithm for non-Gaussianα-stable distribution noise environments based on the correntropy. It is demonstrated by computer simulations that the new proposed algorithm is robust under various noise conditions and presents a very good performance.Finally, an oil pipe leakage alarming system based on generalized correntropy function (GCF) algorithm has been put forward. The Real datas are obtained from the sensors around pipes. After being processed by detection and extraction, the datas are used to testify the abilities of various TDE methods.