Scholarlink

<p>To enhance the recognition accuracy of underwater acoustic target recognition (UATR) via artificial neural networks, a novel UATR approach based on multi-scale features and convolutional residual dense network (CRDNet) is proposed. This paper incorporates a multi-scale convolutional structure into the enhanced ConvNextV2 module and proposes an acoustic feature structure SFbank based on singular value decomposition (SVD). Compared to traditional network frameworks and single acoustic filtering features, this structure demonstrates significant improvements in recognition accuracy, precision, and F1-scores. Experimental validation of the proposed method is conducted on the ShipsEar dataset, achieving a recognition accuracy of 99.08%.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Underwater acoustic target recognition based on multi-scale feature and CRDNet

  • 期刊论文

  • 发表日期：

    2025年9月19日

查看全文
  • Jing Li,
  • Yanru Chen,
  • Xudong Yang,
  • Xinglong Zhang,
  • Lili Zhang,
  • Wei Wei,
  • Pei Yu,
  • Hongxin Tan

摘要

To enhance the recognition accuracy of underwater acoustic target recognition (UATR) via artificial neural networks, a novel UATR approach based on multi-scale features and convolutional residual dense network (CRDNet) is proposed. This paper incorporates a multi-scale convolutional structure into the enhanced ConvNextV2 module and proposes an acoustic feature structure SFbank based on singular value decomposition (SVD). Compared to traditional network frameworks and single acoustic filtering features, this structure demonstrates significant improvements in recognition accuracy, precision, and F1-scores. Experimental validation of the proposed method is conducted on the ShipsEar dataset, achieving a recognition accuracy of 99.08%.