Underwater acoustic signal noise reduction method based on generative adversarial network

An underwater acoustic signal and network technology, applied in the field of underwater acoustic signal, can solve the problems of limited noise reduction ability of low signal-to-noise ratio received signals, insufficient adaptability to complex ocean environments, etc., achieve good noise reduction effect, improve noise reduction performance, adaptable effect

Pending Publication Date: 2021-03-09
PLA STRATEGIC SUPPORT FORCE INFORMATION ENG UNIV PLA SSF IEU
View PDF3 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problems that the existing noise reduction method based on signal processing has limited noise reduction ability for low signal-to-noise ratio received signals, insufficient adaptability to complex marine environment, and strong dependence on domain knowledge and human experience, the present invention provides a A Noise Reduction Method for Underwater Acoustic Signals Based on Generative Adversarial Networks

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Underwater acoustic signal noise reduction method based on generative adversarial network
  • Underwater acoustic signal noise reduction method based on generative adversarial network
  • Underwater acoustic signal noise reduction method based on generative adversarial network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0035] Such as figure 1 As shown, the embodiment of the present invention provides a method for noise reduction of underwater acoustic signals based on generating confrontation networks, including: a training phase and a testing phase; wherein:

[0036]The training phase includes: inputting the noise-containing signal training set into the generator to obtain the signal generated by the generator; splicing the signal generated by the generator and the target signal without noise respectively with the noise-containing signal, and inputting the judgment in sequence In the device; the decision device uses the noise-containing signal as condition information to identify the authenticity of another input signal input at the same time; calculates the error between the judgment result and the real label, and then uses the backpropagation algorithm to complete the generation of the confrontation network Update and optimize the network parameters, and get the trained generator model; ...

Embodiment 2

[0043] On the basis of the above embodiments, the embodiment of the present invention provides a generator construction method, such as figure 2 As shown, the structure of the generator includes: an encoder consisting of seven one-dimensional convolutional layers (Conv1~Conv7), three one-dimensional dilated convolutional layers (D-conv1~D-conv3) and seven one-dimensional A decoder composed of three-dimensional deconvolution layers (T-conv1~T-conv7); among them, each convolution layer in the encoder is connected to each corresponding deconvolution layer in the decoder by using a residual connection; in the In the above encoder, three one-dimensional dilated convolutional layers and decoder, except for the last one-dimensional deconvolution layer in the decoder, the LeakyReLU function of 0.1 is used to activate the output non-linearly after the output of the other layers .

[0044] Specifically, the encoder continuously compresses and reduces the dimensionality of the input si...

Embodiment 3

[0048] On the basis of the above-mentioned embodiments, the embodiment of the present invention provides a construction method of the decider, such as image 3 As shown, the structure of the decision device includes: 3 cascaded step-size convolutional layers (Conv8~Conv10) connected sequentially from the shallow layer to the deep layer, 1 one-dimensional convolutional layer (Conv11) and 1 single-node output A fully connected layer (Fc1); wherein, the three step-size convolutional layers use the same nonlinear activation function as the generator to extract features from the input signal.

[0049] Specifically, the one-dimensional convolution layer in the decision device is a convolution layer with a single convolution kernel, which is used to complete the compression of the feature channel, and connect the obtained one-dimensional feature vector to the fully connected layer, and finally use the The single-node output layer outputs the decision result. and figure 2 same in ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention provides an underwater acoustic signal noise reduction method based on a generative adversarial network. The method comprises a training stage and a testing stage. The training stage comprises the following steps: inputting a noisy signal training set into a generator to obtain a generator generation signal; splicing the generator generation signal and a target signal without noise with a noisy signal respectively, and then inputting the signals into a decision device in sequence; allowing the decision device to take the noisy signal as condition information and performs true andfalse discrimination on another input signal which is input at the same time; and calculating an error between a judgment result and a real label, then updating and optimizing network parameters of the generative adversarial network by utilizing a back propagation algorithm, and obtaining a trained generator model; the testing stage comprises the following steps: inputting a to-be-tested underwater acoustic signal into the trained generator model, and taking the output of the trained generator model as a denoised underwater acoustic signal. The method is suitable for various types of signals,and has a good noise reduction effect on common underwater acoustic communication signals such as MFSK, MPSK, OFDM, LFM and DSSS.

Description

technical field [0001] The present invention relates to the technical field of signal noise reduction, in particular to an underwater acoustic signal based on generative confrontation network. [0002] Noise reduction method. Background technique [0003] Passive detection of underwater acoustic signals is an important research content in the field of underwater acoustic signal processing, and plays an important role in civil applications such as scientific investigation and maritime rescue, as well as military applications such as underwater target monitoring. However, due to the sound absorption of seawater and the complexity and changeability of marine environmental noise, the passive detection ability of underwater acoustic signals is significantly reduced under the conditions of long-distance and fluctuating marine environmental noise. The core of solving this problem is to reduce the noise of the received signal. Researching a noise reduction method that can not only ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G10L21/0208H04B13/02G06N3/04G06N3/08
CPCG10L21/0208H04B13/02G06N3/08G06N3/048G06N3/045
Inventor 李勇斌邵高平曲晶
Owner PLA STRATEGIC SUPPORT FORCE INFORMATION ENG UNIV PLA SSF IEU
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap