Line-spectra extraction of ship-radiated noise based on harmonic wavelet

Abstract

The ship-radiated noise line-spectra contain a large number of characteristic information of the ship, which has great significance for detection and classification of underwater targets. This paper presents a method for extraction of the line-spectra of the ship-radiated noise under strong background noise using harmonic wavelet transformation. Harmonic wavelet is a class of wavelet, which has specific expression for the box-like shape of spectrum and frequency-domain continuous distribution. These features make it more accurate in time-frequency positioning with no frequency leakage in the decomposition process which results in ultra-narrow-band and high resolution. Using harmonic wavelet transformation, the ship-radiated noise is decomposed into separate orthogonal frequency bands which are non-redundant and non-leaking. Based on this, the strong background noise can be suppressed and the line-spectra can be extracted in frequency domain and reconstructed in time domain successfully. Simulation and experimental results show that the harmonic wavelet transform is more effective in the background noise suppression, and its ability for weak line-spectra extraction is superior to Fourier analysis and FIR filter methods.