U-Net structure generative adversarial network and method for underwater acoustic target recognition

• Main Authors: XUE LINGZHI, ZENG XIANGYANG, YANG SHUANG
• Format: Patent
• Language: Chinese; English
• Published: 21.09.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention relates to a U-Net structure generative adversarial network and method for underwater sound target recognition, and the method comprises the steps: building a U-Net structure-based generative adversarial model suitable for underwater target recognition, and enabling the model to well eliminate the overfitting problem of small sample data in a deep learning network, and extracting underwater acoustic features by utilizing jump connection of multi-scale feature extraction, and sending the underwater acoustic features into the generative adversarial network. According to the present invention, the same training set and the same test set are utilized to perform the recognition experiment on the latest underwater target recognition method-based UATC-Densenet method, and the recognition accuracy of the method is compared with the recognition accuracy of the method, such that the recognition rate of the method is superior to the recognition rate of the UATC-Densenet method, and compared with the common