A deep learning method identifies ships by their unique underwater noise. 

 Detecting Ships by Their Underwater Noise: A Deep Learning Approach

Ships make a lot of noise underwater, disturbing marine life and causing interference with underwater communications. This issue, called "Underwater Radiated Noise" (URN), is a mix of broadband noise, tonal sounds, and their harmonics. Recognizing this noise is crucial for adaptive underwater communication and environmental monitoring.

Breaking Down Ship Noise

URN generally comes from various sources on a ship like propellers, engines, and auxiliary equipment. Tonal noise, a consistent, steady sound, often reflects the specific signature of a ship's machinery and can help identify individual vessels.

The New Approach

Researchers Talmon Alexandri and Roee Diamant have developed an innovative way to detect ships using their noise. Their solution revolves around two key techniques:

1.      Deep Learning-Based Detection: Using a custom dataset, a Convolutional Neural Network (CNN) analyzes DEMON (Detection Envelope Modulation On Noise) data. DEMON analysis provides a time-frequency representation of ship sounds, making it easier to spot their patterns.

2.      Narrowband Detector: The detector identifies tonal lines by measuring stability and entropy (a statistical measure of randomness). Using adaptive thresholds, it adjusts detection sensitivity based on the current noise environment.

Test Results

·       Precision and Recall: The new system shows improved precision and recall (finding relevant signals without too many false positives). Even at a Signal-to-Noise Ratio (SNR) as low as 0 dB, it effectively identifies tonal lines.

·       Comparison with Benchmarks: When compared to existing methods, the new approach demonstrates a better balance between precision and recall, detecting the presence of ships more accurately.

Conclusion

The researchers have shown that using deep learning and adaptive detection thresholds provides significant advantages in identifying ships via their unique underwater noise signatures. This advancement opens the door for better marine monitoring, improved adaptive communication, and greater environmental stewardship.

For the full article you are welcome to visit:

https://www.sciencedirect.com/science/article/pii/S1389128624003128

By Talmon Alexandri