The Study Of Underwater Reverberation Modeling And Application Based On Chaos Theory

This thesis is one part of an underwater acoustic project. How to detect the weak echo signal under the strong reverberation background in this project is a key problem to solve. Some of the related theoretical and experimental achievement will be covered in the thesis.Oceanic reverberation, the fiercest interference to the active sonar, should be suppressed by the sonar signal processing. It has been a long time that the reverberation is considered as a stochastic process and the approaches to the oceanic reverberation problem are based on the statistic theory. In recent years, some researches by the nonlinear dynamics on some physical systems discover the deterministic natures that mismatch their disorder behavior. This is chaos that possesses regular characteristics but exhibit irregular observation to the system like random. Under the illumination remark of chaos, this thesis raises the idea of modeling oceanic reverberation as an observation of a nonlinear deterministic system by the dynamics. And some groping researches about signal processing are made on this basis of the chaotic model.Firstly, chaos and fractal theory is introduced, such as the definition of the chaos, Lyapunov exponent, fractal dimension, strange attracter and so on, which are the basic theories throughout this thesis. Secondly, Brief explanation of the experiment carried out in the laboratory and outfield is provided, with the source of the data used in the thesis in the appendix.The state space reconstruction that is realized by delaying coordinate is introduced for studying the phase space structure of the reverberation time series. In order to compute the reconstruction parameters, lag and embedding dimension, of the reverberation, average mutual information and Cao's method is utilized respectively. The analytic results verify that the phase space structure of the reverberation has low-dimension. Reconstruction quality standard makes the study of the reconstruction distortion more convenient. In this paper, Q_M the function of eigenvalues of the distortion matrix is used to assess the reconstruction quality. And estimation method about the reconstruction window is found by analyzing the numerical value of Q_M . It is the prerequisite of this thesis that if the reverberation series have chaotic property. Moreover, this problem has not been solved drastically in the reverberation-free research. The maximal Lyapunov exponent of same typical reverberation is calculated, the result indicates that the maximal Lyapunov exponent of the reverberation is positive. After that, chaotic forecast model is built using the Volterra adaptive filter, which carries out the short period forecast of the reverberation through one-step and multi-steps. The performance comparison between Volterra forecast model, local forecast model and AR forecast model reveals that the reverberation is more suitable for the deterministic model.Signal chaotic detection under reverberation background is one of the emphases of the paper. Chaotic model forecast error is used to detect the echo under low signal-reverberation -ratio. The detection ability is studied with different signal-reverberation-ratio. It can detect the echo under the reverberation background if the signal-reverberation-ratio greater than -3dB, and its ability is better than matched filter.Minimum phase space volume (MPSV) is one of the methods to estimate the spectrum of the signals under the reverberation background. This method is used to estimate the frequency of simulated and real signals. Compared with standard least squares estimation, the MPSV estimator is shown to be more accurate and requires only a short data record.According to the demand of the engineering project, another emphasis of this paper is echo waveform extract under the reverberation. Neighborhood method is proposed to solve this problem. This method projects the reverberation to the neighborhood space, and the signal is reserved in the orthogonal space. The efficiency of this method is further investigated using real-life reverberation and echo. The results show that this method performs better when the signal-reverberation-ratio is greater than 3dB.