A Novel Deep-Learning Method with Channel Attention Mechanism for Underwater Target Recognition

Area of Science:

• Underwater Acoustics
• Signal Processing
• Machine Learning

Background:

• Underwater acoustic recognition relies on spectral feature extraction.
• Doppler shifts from target movement complicate recognition.
• Existing methods struggle with diverse working conditions.

Purpose of the Study:

• To propose a novel deep learning approach for enhanced underwater acoustic target recognition.
• To address challenges posed by Doppler frequency variations.
• To improve recognition accuracy across different operational scenarios.

Main Methods:

• Developed a ResNet (Residual Network) for deep spectral feature extraction.
• Integrated a channel attention mechanism into ResNet (camResNet) to enhance stable spectral features.
• Employed a one-dimensional convolutional neural network for feature classification.

Main Results:

• Achieved a top recognition accuracy of 98.2% on challenging datasets.
• Demonstrated superior performance compared to other recognition approaches.
• Showcased improved robustness over standard ResNet with channel attention for varied working conditions.

Conclusions:

• The proposed camResNet approach effectively enhances underwater acoustic target recognition.
• The method accurately identifies targets despite Doppler shifts and diverse working conditions.
• This deep learning strategy offers a promising solution for robust underwater acoustic identification.