Method for identifying target in water

Abstract

The invention discloses a method for identifying a target in water. Wavelet packet decomposition threshold denoising is combined with component difference optimization to finish preprocessing of the radiation noise signal; the problem that ocean background noise affects radiation noise time-frequency feature extraction and interference classification and identification is solved; mapping from a one-dimensional sequence signal to a two-dimensional space is achieved by adopting a wavelet transformation method; the problem that a single time domain or frequency domain characteristic representation signal is not comprehensive is avoided; two-dimensional variational mode decomposition is carried out on the time-frequency characteristics; the interference problem of a two-dimensional space noisesignal is solved; feature optimization is carried out on the obtained intrinsic mode component and a signal is constructed according to the feature optimization, characteristic enhancement of time-frequency characteristics is achieved, two-dimensional variation modal decomposition is carried out through employing an edge mirror image method, a signal oscillation problem caused by an edge effect is avoided, gradient descent training is carried out on a small sample data set to update parameters of a classification discriminator of the deep neural network, and the characteristic extraction model is made to have excellent generalization capability.

Description

technical field [0001] The invention belongs to the field of underwater acoustic signal processing, and in particular relates to an underwater target recognition method. Background technique [0002] The radiated noise signal of underwater acoustic targets is characterized by numerous and concentrated sound sources, diverse spectrum components, and high radiated noise intensity. However, the marine environmental noise caused by ocean turbulence and seawater static pressure effects, etc., its spectral components cover the entire frequency band and change continuously with various factors, making underwater acoustic signals very complicated. The research on the classification and identification of underwater acoustic target radiation noise signals mainly focuses on machine learning. The classification effect of traditional machine learning methods depends on the quality of hand-designed features, which has a strong subjective prior; deep learning relies on the model itself. T...

AI Technical Summary

Problems solved by technology

At present, most underwater target recognition extracts feature information through a single transformation in the time domain or frequency domain. The single-dimensional signal feature is a "static" index, and the expression information is not comprehensive. After the feature transformation, the original expression ability of the signal is lost, which is not easy. It is intuitively understood by people that the lack of semantic information makes it difficult for the classifier designed as a training sample to have high accuracy; realizing underwater target recognition based on deep learning requires a large amount of training data, and the model training cycle is long, which does not meet the high efficiency of underwater target recognition requirements

Images