Automatic recognition method of non-cooperative underwater targets based on Gaussian mixture model

A Gaussian mixture model, water target technology, applied in character and pattern recognition, instruments, computing and other directions, can solve problems such as poor practicability, reduce data volume, improve recognition performance and robustness, reduce label information and data volume the effect of the request

CN111027453BActive Publication Date: 2022-05-17NORTHWESTERN POLYTECHNICAL UNIV
9 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Publication Date
2022-05-17

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a non-cooperative underwater target automatic identification method based on a Gaussian mixture model, which is used to solve the technical problem of poor practicability of the existing non-cooperative underwater target identification method. The technical solution is to use the known in-set data structure to extract the Mel frequency cepstral coefficients, which can describe the nonlinear characteristics of human hearing, from the perspective of the way of recognition inside and outside the set when facing targets of unknown categories, that is, MFCC coefficients , train the Gaussian mixture model, select the appropriate threshold, and thus construct the target recognition system, by substituting the samples that are difficult to obtain label information into the target recognition system, to distinguish whether they belong to the samples in the known sample set, and judge their categories to achieve Preliminary judgment is made on the category of non-cooperative targets, and at the same time, the requirements for label information and data volume of sample data are reduced, and the practicability is good.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to an underwater target recognition method, in particular to a non-cooperative underwater target automatic recognition method based on a Gaussian mixture model. Background technique

[0002] The document "Application Progress of Deep Learning in Passive Recognition of Underwater Targets, Signal Processing, 2019, Vol35(9), p1460-1475" discloses a method for passive recognition of underwater targets based on deep learning. After performing the steps of typical pattern classification and recognition systems such as preprocessing and feature extraction, a specific deep neural network structure is used to realize the classifier design and classifier link, and a certain number of samples are used to complete the model training of the deep neural network, or directly use the depth The neural network has the characteristics of good feature learning ability, and the use of deep neural network weakens or completely replaces the key feature ...

Examples

Embodiment Construction

[0044] refer to Figure 1-4 .

[0045]The data set used in this example contains 3 types of underwater acoustic targets, 15 segments of sound are collected for each type, and the length of each segment of sound is about 6 seconds. The frequency is 8000Hz. During the test, class I and class II data are used as in-set data, and class III is used as out-of-set data. Select a part of the type I and type II data as the training set to train and build the GMM model, and the rest of the type I and type II data and type III data are used as the test set to test the trained GMM model.

[0046] The specific steps of the non-cooperative underwater target automatic recognition method based on the Gaussian mixture model of the present invention are as follows:

[0047] Step 1: Preprocessing the training samples observed in the data set, including three parts: pre-emphasis, framing, and windowing; the training samples are underwater acoustic target data.

[0048] Using MATLAB as the pla...