Distinguishing multiple surface ships using one acoustic vector sensor based on a convolutional neural network.

Area of Science:

• Acoustics
• Signal Processing
• Machine Learning

Background:

• Accurate direction of arrival (DOA) estimation is crucial for identifying and tracking multiple acoustic sources.
• Traditional methods face challenges in complex underwater environments with multiple interfering signals.
• Acoustic vector sensors offer richer spatial information compared to traditional pressure sensors.

Purpose of the Study:

• To propose a novel DOA estimation method using a convolutional neural network (CNN) and acoustic vector sensor data.
• To enhance the capability of distinguishing multiple surface ships in a specific frequency band.
• To leverage machine learning to improve DOA estimation performance by learning acoustic propagation characteristics.

Main Methods:

• Utilizing an acoustic vector sensor to capture both pressure and particle velocity data.
• Inputting the cross-spectrum of pressure and particle velocity into a CNN model.
• Training the CNN using simulated data generated with an acoustic propagation model under varied environmental and source parameters.
• Experimental validation using real-world acoustic data.

Main Results:

• The CNN effectively learns acoustic propagation characteristics, leading to improved multisource distinguishing performance.
• The proposed method demonstrates accurate DOA estimation for multiple surface ships.
• Experimental results confirm the practical applicability and effectiveness of the developed technique.

Conclusions:

• The CNN-based DOA estimation method using acoustic vector sensors provides a robust solution for multisource identification.
• Learning acoustic propagation patterns significantly enhances the performance in distinguishing multiple surface ships.
• The method shows promise for real-world applications in underwater acoustics and surveillance.