The Study Of Multipath Time Delay Estimation Algorithm In Alpha Stable Noises

Multipath time delay is of great value in many areas. Research on time delay promotes the development of digital signal processing, information theory and signal detection.etc. Thus, time delay estimation not only of theoretical significance, but also of practical value. In time delay study, we usually use Gaussian Distribution Model to describe signal and noise in reality. But the practical application of underwater acoustic signal, low-frequency atmospheric noise and man-made noise signal has a spike characterized by non-Gaussian signal and noise. If still using the traditional Guassian distribution model to describe these signal and noise, the effect of estimation function will degradation and fail, since signal and noise would not be a good match with model. In order to describe the characteristics of the signal with spikes and noise accurately, this thesis uses Alpha stable distribution.Alpha stable distribution is the only distribution which can satisfy central limit theorem and able to describe the non-Guassian noise more accurately. It is able to maintain the mechanism and propagation conditions in natural noise processing. Since Alpha stable distribution of stability, it can describe actual dates accurately.Firstly, this thesis uses the theoretical analysis and function analysis to analyze time delay estimation methods based on second-order statistics, such as correlation, generalized correlation, second correlation and spectroscopy and so on. Since Alpha stable distribution is no finite second order statistical moments, it has limitations using traditional time delay estimation method based on second-order statistics.Therefore, this thesis mainly study multipath time delay estimation based on fractional lower order statistics, such as EM multipath time delay and WR multipath time delay estimation. The author will analyze the two methods above, and make a compare of their between each other. Then this thesis will improve the shortcoming, namely generalizing EM algorithm and WR algorithm, so that enable them can overcome the drawbacks of traditional methods. To expand its areas of application.