Research Of Mechanical Chirp Signal Detection And Time Delay Estimation

Reflected Chirp mechanical wave exploration is a very important exploration method in many science and engineering domains such as radar detection, controlled vibrator exploration of underground stratum structure, sonar detection of object underwater, and so on, most of which has belonged to high technology science and engineering. The level of reflected exploration lies on three factors including level of the staff, accuracy of the instruments and level of signal process, whereas the main difficult in signal process if how to depress the environmental noise in the received signals, especially the noise from the movement of machines such as vehicle.The traditional echo detection arithmetic is correlation arithmetic, in which the sent signal and received signal are cross-correlated, and impulses will appear in the positions of echoes. Nowadays more and more exploration is performed in environments of strong mechanical noise such as city, factory and mine, and so forth, where the received signals will be mixed with strong mechanical noise. Correlation arithmetic is one of the globally used methods of signal detection, and it is convenient, practical and simple, but sometimes its ability of depressing noise is not optimal, and its differentiation is not accurate.In the analysis of mechanical noise sampled from urban environment, many samples of noise present non-Gauss probability distribution. It can be proved theoretically that correlation operation in the known signal detection is equal to signal detection in Gauss noise, so correlation operation is fit for depressing Gauss noise. High order statistics is a new mathematics tool in signal process with strong functions, it can not only depress non-Gauss colored noise, but also can it screen Gauss colored noise automatically, so it can be used to depress urban mechanical noise. Correlation and three-order statistics arithmetic has been developed systematically in stationary random signal process, while STFT and wavelet transform arithmetic are based on time-frequency distribution, which can be sorted to non-stationary random signal analysis arithmetic. But there is no optimal stationary signal in the nature, the non-stationary is absolute and the stationary is related. Many random signals can be analyzed as both stationary random signal and non-stationary random signal, so wavelet transform arithmetic is also tried in detection of echoes and time delay estimation under the urban mechanical noise. This paper includes totally eight chapters, in which the arithmetic of correlation, three-order cumulant, three-order spectrum, continuous wavelet transform(CWT), discrete wavelet transform(DWT) and wavelet package are all researched, and are all used in Chirp signal detection and time delay estimation. The practical urban mechanical noise is used in the numerical test to compare the ability of depressing noise and detection resolution.The first chapter discussed the characteristics of Chirp signal and the principle of reflected Chirp exploration, and talked about the development history and current state.The second chapter analyzed the sampled noise used in the experiments, researched its frequency distribution, probability distribution and attenuation characteristic, offer an important reference to the Chirp signal detection and time delay estimation. It was found that the frequency spectrum of urban mechanical noise lies mainly from 0 to 50Hz, and part of the noise samples take on different degrees of Gauss probability distributions, but the others take on probability distributions obviously different from Gauss probability distribution. The third chapter discussed the mathematics theoretical base of stationary random signal analysis, including random variable and random course, the definition and related properties of high-order statistics, the statistical characteristic of Gauss random course. A mathematical prove was given that the probability characteristic of Gauss random course can be overall described by two-order statistics and their high-order...