Scholarlink

The spatial spectrum estimation is a significant research direction in array signal processing filed, which is crucial for underwater acoustic target orientation. The spatial spectrum estimation is mainly carried out around the covariance matrix. However, when faced with complex environment, the effective information of covariance matrix is blurred. To address the issue of poor spatial spectrum estimation accuracy at low Signal-to-Noise Ratio (SNR) and short signal frame length conditions, 3-Dimensional Physical Information Enhancement Neural Network (3-D PIENN) method is proposed for Underwater Acoustic Target Spatial Spectrum Estimation (UATSSE). The input information of 3-D PIENN is the covariance matrix of the signal frequency spectrum. The frequency feature information which is consistent with the signal frequency is extracted by constructing 3-D PIENN. And the spatial and correlation feature information of the signal is preserved and enhanced to great extent. The experimental results show that compared with MVDR, MUSIC and Res-CNN, the proposed method has the highest spatial spectrum estimation accuracy at low SNR and short signal frame length conditions.

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

Spatial Spectrum Estimation of Underwater Acoustic Target Based on 3-D Physical Information Enhancement Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Honghui Yang,
  • Qiang Guo,
  • Junhao Li,
  • Kaifeng Zheng,
  • Jing He

摘要

The spatial spectrum estimation is a significant research direction in array signal processing filed, which is crucial for underwater acoustic target orientation. The spatial spectrum estimation is mainly carried out around the covariance matrix. However, when faced with complex environment, the effective information of covariance matrix is blurred. To address the issue of poor spatial spectrum estimation accuracy at low Signal-to-Noise Ratio (SNR) and short signal frame length conditions, 3-Dimensional Physical Information Enhancement Neural Network (3-D PIENN) method is proposed for Underwater Acoustic Target Spatial Spectrum Estimation (UATSSE). The input information of 3-D PIENN is the covariance matrix of the signal frequency spectrum. The frequency feature information which is consistent with the signal frequency is extracted by constructing 3-D PIENN. And the spatial and correlation feature information of the signal is preserved and enhanced to great extent. The experimental results show that compared with MVDR, MUSIC and Res-CNN, the proposed method has the highest spatial spectrum estimation accuracy at low SNR and short signal frame length conditions.