Underwater Acoustic Target Recognition Based on Attention Residual Network

Area of Science:

• Marine acoustics
• Signal processing
• Machine learning

Background:

• Underwater acoustic target recognition faces challenges due to limited labeled data, complex marine environments, and background noise interference.
• Accurate identification of underwater acoustic targets, such as ship-radiated noise, is crucial for various applications.

Purpose of the Study:

• To propose and evaluate an attention-based residual network (AResnet) for robust underwater acoustic target recognition.
• To enhance the accuracy and reliability of ship-radiated noise identification in diverse marine conditions.

Main Methods:

• Utilized a residual network for deep feature extraction from 3D fusion features.
• Incorporated a channel attention module to enhance feature representation.
• Employed joint supervision of cross-entropy and central loss functions for classification.
• Applied pre-trained AResnet for feature extraction and fine-tuning on new datasets.

Main Results:

• Achieved 99% average recognition accuracy on the DeepShip dataset.
• Attained 98% average recognition accuracy on the ShipsEar dataset after fine-tuning.
• Demonstrated superior performance compared to existing methods.

Conclusions:

• The proposed AResnet method significantly improves underwater acoustic target recognition accuracy.
• AResnet offers a robust solution for identifying ship-radiated noise in various environments.
• Fine-tuning pre-trained AResnet enables effective application to new underwater acoustic datasets.