Detection And Feature Extraction Of Radiated Noise From High Speed Underwater Vehicle

With the development of modern science and technology, the abilities of a high speed underwater vehicle in the speed, range, depth and hidden increase significantly. Early-warning and feature extraction of the underwater targets is an important research project for the passive sonar's signal processing. Radiated noise from a high-speed underwater vehicle includes a variety of signals, such as water entry sound, navigating signal and homing signal. In this paper, basic research and analysis is done using a variety of processing methods in various angles. The thesis is divided into three parts:analysis and modeling of radiated noise from high speed underwater vehicle, underwater transient signal detection techniques and feature extraction techniques of high speed underwater vehicle's radiated noise.Firstly, according to the basic characteristics of radiated noise from a high-speed underwater vehicle, establish mathematical model in keeping with practicality. Simulate the water entry sound, navigating signal and homing signal, and the results are closer to the real signals.Based on the short term energy detector, the Power Law detector and the Page Test detector used in underwater transient signal detection, propose the WP-STED, the WP-Power Law detector and the WP-Page Test detector. With the wavelet packet analysis, the SNR of the input signal could be improved and then utilized those detectors. Simulations are given, and the performance of those methods can be superior to those traditional detectors. To the water entry sound and the launch tube covering signal, the detection results prove the valid of those algorithm.Study on the feature extraction techniques of radiated noise from the high speed underwater vehicle. Combined the time-frequency analysis method with image processing, utilize the STFT-RT approach for feature extraction of underwater moving target's radiated noise. Based on increasing the SNR of time frequency image, the Radon transform is applied to reduce the feature dimension. Then the characteristics of frequency variation on variable-speed navigation and blade frequency on stationary navigation could be calculated automatically. Introduce the DEMON spectrum analysis method and the adaptive line enhancement device, and get the propeller shaft frequency and harmonic frequency of navigating signal. Reference the region growing image processing technology, search along the time axis of the time-frequency spectrum, to achieve spectral line tracking of navigating radiated noise. Simulations verify the effectiveness of those algorithms, a variety of signal features extracted from the actual data are conducive to classification and identification.This study aims at overall understanding and extracting the high speed underwater vehicle radiated noise feature, in order to provide support for improving sonar performance of the alarm and identification.