Analysis Of Ship-radiated Noise And Reconstruction Of Waveform Based On Bispectrum

Higher-order Statistics (HOS) is the primary analysis tools which can be used to analyze noneasual system, nonminimum phase system and nongaussian signal. It possesses plenty of advantages in signal detection, harmonic retrieval and feature extraction. On the bases of theory analysis, combining whth National Defense Predictive Projects, the usage of HOS in signal processing domain, especially in features analysis based on ship-radiated noise, is explored in the paper.First, the mechanism of generation of ship-radiated noise, and then, the definition and properties of HOS are introduced. We especially notice that HOS is insensitive to Gaussian process. This is the theoretical basis for us to use it in signal detection and estimation. Bispectrum is the method widely used in signal processing for its less calculations and higher resolution. Using the method of harmonic reconstruction based on non-parameter bispectrum estimation, the line spectrum components of ship-radiated noise can be reconstructed precisely. In addition, the envelope of ship noise waveform can be remodulated utilizing Hubert transform, and the spectrum components with line rotation frequency of propeller shaft are extracted effectively by bispectrum analysis. Computer simulations and ship-radiated noise data show that the method can effectively extract harmonic features and shaft rotation frequency.In this dissertation, the harmonic reconstruction based on parameter bispectrum estimation is also discussed. Comparing with non-parameter bispectrum estimation, parameter bispectrum estimation can achieve higher resolution, less number of data are required, but it is very important to select the correct model order.At last, the reconstruction method of continuous spectrum components is introduced in this dissertation, and then the waveform reconstruction of ship-radiated noise is completed with superimposed line components of the noise.