Research On Passive Sonar Signal Modeling Of Underwater Moving Target

In the field of underwater acoustic processing, the process and analysis of passive sonar signal of underwater moving target has always been the research hotspot in this field, and it is also the important link and key technology in the underwater acoustic warfare and torpedo warning. However, the uncertainty of water environment and sound propagation conditions make the signal has the time-varying character, the actual measured passive sonar signal of underwater moving target is very weak and includes lots of background noise and interference, so that it is difficult to extract the effective component. Meanwhile, getting the actual signal data needs specialized equipment and personnel, and it is time-consuming; expensive and the data is security. All these factors restrict the analysis and application of passive sonar signal (such as the maritime forces training, the scheme of target commend system’s argumentation, as well as the sonar simulation training systems in the laboratory study of the test phase). Therefore, the simulation of passive sonar signal of underwater moving target has important practical value and military significance.This paper analyzed the characteristics of the underwater moving targets radiated noise (original signal for short in the following part), including the mathematical and statistical properties and generating mechanism of original signal. On base of this, the research simulates the liner spectrum and continuous spectrum containing in the original signal. Besides, aiming at many influencing factors when the original signal propagates in the complex ocean environment, the research stimulated the passive sonar received signal (target signal for short in the following part).Due to the neural network has self-organization, self-adaption and multi-threaded parallel processing features, as well as its good approximate for nonlinear system, this research explores the application of neural network to simulate the process of original signal propagating in the nonlinear channel and makes innovative research on the RBF neural network algorithm. On the basis of in-depth discussion of K-means clustering algorithm and the classic FCM algorithm, it proposes the improved FCM algorithm which has better generalization ability and self-adaption. By using the improved algorithm to make clustering analysis on the signal data to determine the structure and parameters of the RBF neural network, this research verify the effectiveness of the improved algorithm for the nonlinear system by approximating and forecasting simulation on nonlinear fuction. Finally, by using the improved algorithm the research modelings the nonlinear processes of the propagation of original signal to the passive sonar; stimulates the passive sonar signal and its features of time domain. Besides it uses the correlation coefficient and error index as evaluation indexes to make similarity compatision on the model output data and target signal, the result prove that the improved algorithm is better than the classic algorithms achieving the purpose of underwater passive sonar signal of the underwater moving target modeling.