Research On Underwater Acoustic Signal Based On Nonlinear Theory

With the development of nonlinear science,including chaotic and fractal theory,using nonlinear signal processing to study the signal's nonlinear properties is get more and more attention. Traditional signal processing decomposes the signal into two poles:deterministic part and stochastic part. However there are some signals different from the two poles,named chaotic signals,which are generated by a determinate system but acting similar to stochastic signal and can not be forecasted in long term.As a kind of underwater acoustic,the ship-radiated noise is generated by a very complex nonlinear system. Many recent researches show that underwater acoustic,including ship-radiated noise,is chaotic. Therefore applying the nonlinear signal processing to the underwater acoustic to study the target's nonlinear properties is a way of modern underwater acoustic processing.Present nonlinear signal processing researches on ship-radiated noise are focus on detection,feature extraction and target recognition. There is almost no one to study the nonlinear dynamic modeling of ship-radiated noise. Based on the phase space reconstruction theory,this dissertation studies the chaotic time series' dynamic modeling technology using the Volterra Series,and brings out a method to get the model's parameters through the forecasting of the time series,and finally,applying to the ship-radiated noise.The main endeavor and contributions of the dissertation are as follows:1. Introduced the principles and methods to forecasting chaotic time series,and using the Volterra Series to model nonlinear dynamic systems. A global forecasting method of chaotic time series based on the Volterra Series is put forward.2. Discussed the numerical calculating quantity of the Volterra Series,and put forward a Separable Volterra Series Model based on the representing of Volterra Series with tensor production,which evidently reduced the numerical calculating quantity and unproved the performance.3. Studied the relationship between the forecasting and the dynamic systems. The result shows that the forecasting model of a chaotic time series can be used to evaluate the original dynamic systems. So we can estimate the model's parameters of a dynamic system by forecasting.4. Studied the selection of the dimension,delay-time and order of Volterra Series. Tested the Volterra Series' anti-noise performance. Bring out the Local Projection method to reduce additional noise of chaotic time series in phase space.5. Bring out a new way to determine whether a dynamic system is chaotic by using the model of the system.6. For phase space reconstruction of ship-radiated noise,using auto correlation function and mutual information to select the delay-time,and using False Nearest Neighbors method to select the embedding dimension.7. Applying Volterra Series to nonlinear dynamic system modeling of ship-radiated noise. The result shows that Volterra Series is suitable for modeling the ship-radiated noise. In the sense of Volterra Series model,the ship-radiated noise is chaotic,but contains periodic or pseudo-periodic signal in main part.