Submarine sound signal noise reduction method

Abstract

The invention provides a submarine sound signal noise reduction method. The method comprises steps that firstly, a submarine sound signal is inputted to decompose into multiple modal components; secondly, the multiple modal components acquired through decomposition are transformed to a frequency domain for processing; thirdly, a threshold function is set according to noise distribution characteristics of the submarine sound signal to filter noise; fourthly, the multiple modal components after processing are transformed to a time domain; and lastly, reconstruction of the submarine sound signal after noise reduction is carried out. The method is advantaged in that compared with a method in the prior art, noise is effectively eliminated, through decomposing the submarine sound signal into a series of modal components, identification of noise in the modal components is facilitated, and subsequent processing is convenient; compared with a common spectral subtraction method, the method has an advantage in processing non-stable signals, the physical meaning of the signal itself can be better presented, moreover, through designing the threshold function, the modal components are transformed to the frequency domain to carry out filtering processing, and energy statistic characteristics of the noise and the signal are fully utilized; compared with a common wavelet noise reduction method, the method has higher noise reduction performance.

Description

technical field [0001] The invention belongs to the field of acoustic signal processing, and relates to underwater acoustic signal noise reduction technology, in particular to a threshold value noise reduction method. Background technique [0002] When the underwater acoustic signal propagates in seawater, due to the interference of other targets, complex and changeable ocean environment, and the influence of instruments, the receiving sensor receives a non-stationary signal with a very low signal-to-noise ratio. Therefore, in order to extract and classify the target features at the receiving end of the underwater acoustic signal, it is first necessary to filter the collected original signal. [0003] The underwater acoustic signal superimposed with ocean noise is a time-varying non-stationary signal without a uniform waveform, so it is difficult to design a suitable matched filter. Traditionally, there are some noise reduction methods based on Fourier transform, such as sp...

AI Technical Summary

Problems solved by technology

However, wavelet denoising has disadvantages such as Gibbs effect and low reconstruction accuracy, and because the wavelet base is preset, it cannot represent the real signal in reality well.

Independent Component Analysis (ICA) is a signal processing technology that has emerged in recent years. If the source signals are independent, ICA can decompose the received mixed signal into mutually independent components, and the separated components The component is the source signal, but there is also a restriction that each independent component must be non-Gaussian distributed

Images