Research On Ship Noise Analysis Method Based On Signal Enhancement And Feature Extraction

With the development of science and technology,human activities on the sea are gradually increasing.Ship noise is an important component of ocean noise.There is an increasing demand for analysis and detection of ship noise.Ship radiated noise is a significant basis for target detection,tracking and positioning.However,due to the complex and volatile marine environment,there are various kinds of noise in the ocean,and the identification of ship noise is more difficult.The demand for robust ship detection systems is increasing rapidly.Research on ship noise signal enhancement,feature extraction and identification methods is of great significance for marine safety and national defense construction.This thesis analyzes the mechanism and characteristics of ship radiated noise and establishes the ship noise model,and proposes a combination of Gammachirp Filter Bank(GCFB)and Resonance-based Sparsity Signal Decomposition(RSSD).This method is called GCFB and RSSD Combination Technology(GRCT),combines GCFB with RSSD to achieve the weak signal enhancement of ship noise.We discusse the performance of different time-frequency analysis methods,and conclude that the performance of Polynomial Chirplet Transform(PCT)is optimal.A PCT time-frequency aggregation method based on maximum pooling(MP TFFPCT)is proposed.This method is used to analyze ship noise,extract time-frequency features,and fuse with auditory features to obtain a ship noise detection system with better recognition performance.The feasibility and robustness of the system are verified by real ship noise signals and a large amount of simulation data.The main works of this paper are listed as follows:(1)The ship noise model is established,including continuous spectrum and line spectrum model.The common underwater acoustic channel model is analyzed,and a simulation method of ship noise signal based on sonar equation is proposed.The simulation signal obtained by the model is compared with the real signal to verify the validity of the model,which provides the basis for subsequent analysis.(2)GRCT is proposed by combining GCFB with RSSD for ship noise signal enhancement.GCFB is used to enhance the interesting harmonic components adaptively and reduce the noise interference.Then RSSD algorithm is used to extract the high-resonance component of the signal,filter out the in-band noise and transient interference,so as to achieve the purpose of signal enhancement and improve the recognition performance.(3)A variety of time-frequency analysis methods are introduced.Quantitative indicators are used to evaluate the performance of each method,and verify that PCT performance is optimal.The MP TFFPCT method is proposed to analyze the time-frequency distribution of ship noise signals.This method is used to extract the features of different types of ship noise signals,and obtain time-frequency features,which are fused with auditory features.The fusion features have better recognition performance under low signal-to-noise ratio.