Robust Time Delay Estimation Algorithms In The Presence Of Alpha Stable Distribution Noise

Time delay estimation is one of the most significant parameter estimations in practical engineering applications such as source localization,speech,control,astronomy,and medical auxiliary diagnosis.Many time delay estimation algorithms depend on the Gaussian noise assumption which is usually acceptable in many situations.But not for the cases of wireless communication,radar,underwater sound and biomedical where noise is impulsive.In this situation,it is more proper to depict the noise through alpha-stable distribution other Gaussian distribution.Many of the existing algorithms for traditional alpha-stable distribution signal processing are based on the fractional lower order statistics(FLOS),which leads to their limits.For most cases,the order p should be smaller than the exponent characteristic of alpha-stable distribution which depends heavily on the a priori knowledge of the noise and it is difficult to obtain.Furthermore,the FLOS does not have the second-order statistics and therefore,some information is lost.In order to achieve better accuracy and robustness of the time delay estimation algorithms,the research is carried out in following three aspects:the integer TDE,the fractional TDE and the joint estimation of the time delay and Doppler frequency shift.The main research work and innovation could be summarized as follows.(1)In order to solve the integer time delay estimation problem under alpha-stable distribution,the weighted correntropy spectrum TDE method is proposed in this paper.It introduced the concept of the correntropy spectrum as well as the whitening procedure,through which we achieve better robustness and lower calculation complexity.The proposed algorithm shows its superiority for time delay estimation under lower generalized signal-to-noise ratio(GSNR).(2)In order to solve the TDE problem for higher accuracy and lower sampling rates,two fractional TDE methods under alpha-stable distribution noise are proposed in this paper.It proved the effectiveness of correntropy induced metric(CIM)under alpha-stable distribution noise and then proposed the fractional TDE algorithm based on the minimum CIM and the Farrow structure.Although this method could achieve accurate estimation result for wide-banded signals,it performance degrades for narrow-banded signals.Therefore,a novel algorithm based on the maximum correntropy criterion(MCC)and the Lagrange fractional filter is proposed,which has better results for both Gaussian and non-Gaussian noise environment as well as different signals of various band-width and modulation.(3)In order to solve the joint time delay and Doppler frequency shift estimation problem under both impulsive noise and co-channel interference,this paper proposed the concept of cyclic-correlation based on sigmoid transformation and then proposed the joint estimation algorithm based on sigmoid transformation cyclic ambiguity function.The novel algorithm applied the nonlinear sigmoid transformation to heighten robustness to impulsive noise and achieved better accuracy under lower GSNR.Furthermore,this paper proposed multiple adjustable sigmoid transformation ambiguity functions,through which it can obtain more information of time delay and Doppler frequency shift and get better noise resistance ability.The proposed joint estimation algorithm based on the generalized sigmoid transformation cyclic ambiguity function functions works well under low GSNR and co-channel interference environment.