Robust shallow water reverberation reduction methods based on low-rank and sparsity decomposition

Area of Science:

• Signal Processing
• Sonar Systems
• Array Signal Processing

Background:

• Low-rank and sparsity decomposition effectively reduces reverberation in multi-ping detection.
• Performance degrades with highly sparse reverberation or stationary targets due to ambiguous signal characteristics.

Purpose of the Study:

• To derive the identifiability condition (IC) for low-rank and sparse matrix decomposition.
• To propose methods for improving reverberation reduction and target detection in challenging scenarios.

Main Methods:

• Derived the identifiability condition (IC) from random orthogonal and sparsity models.
• Proposed sparsity compensation for low-rank matrices to address false alarm probability inflation (FAPI).
• Proposed increasing sparse matrix dimensions to manage detection probability shrinkage.

Main Results:

• Sparsity compensation effectively eliminates FAPI by increasing the sparse coefficient of the low-rank matrix to 0.30.
• Stationary targets can be detected with a high dimension (large ping number) of the sparse matrix.
• Validated robust reverberation reduction performance through simulations and field experiments.

Conclusions:

• The proposed methods ensure meaningful decomposition and robust performance in reverberation reduction.
• Identifiability conditions are crucial for reliable signal decomposition in sonar.
• The study provides practical solutions for improving sonar detection capabilities.