Active target echo simulation method based on diffusion model

Through the active target echo simulation method based on the diffusion model, the problem of difficulty in underwater target echo simulation is solved, the accurate simulation of underwater target echo signals and the expansion of sample libraries are achieved, and the accuracy of target recognition is improved.

CN120197483APending Publication Date: 2025-06-24THE 715TH RES INST OF CHINA SHIPBUILDING IND CORP +1
0 Cites 1 Cited by

Patent Information

Application Number
CN202510268604.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-07
Publication Date
2025-06-24

AI Technical Summary

Technical Problem

The prior art is difficult to accurately simulate the propagation of sound waves underwater and the echo characteristics after interacting with the target, which makes it difficult to simulate the echo of the target underwater.

Method used

The active target echo simulation method based on the diffusion model is adopted, and the diffusion model training data set is constructed, and the feature extraction framework is constructed using the UNetV0 network, and the iteration of noise added cycles and gradient back-passing is performed to optimize the model parameters, generate target information and obtain simulation samples.

Benefits of technology

The precise simulation of the underwater target echo signal is achieved, the target sample library is expanded, and the target recognition accuracy is improved.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides an active target echo simulation method based on a diffusion model, and the method comprises the following steps: 1, constructing a diffusion model training data set, carrying out the preprocessing and marking of echoes containing target information, and constructing a trainable data set; step 2, constructing a diffusion model, and constructing a feature extraction framework based on a UNetV0 network for extracting and reconstructing target information in a spectrogram; 3, adding noise loop iteration in the training process, and performing gradient return to optimize model parameters; step 4, when the diffusion model is trained, using mean square error loss to minimize noise distribution applied in a forward process and noise distribution predicted in a backward process; and step 5, in a sample generation process, generating a noise through the set random function, and then obtaining a target sample through noise data sampling iteration. The method has the advantages that the signal simulation method is combined with an intelligent recognition method, simulation sample expansion can be achieved, and the target recognition accuracy is improved.
Need to check novelty before this filing date? Find Prior Art