Ship Radiated Noise Recognition Technology Based on ML-DS Decision Fusion

Area of Science:

• Underwater acoustics
• Signal processing
• Machine learning

Background:

• Ship radiated noise is crucial for identifying underwater targets.
• Interference noise degrades recognition accuracy in traditional methods.
• Accurate identification is vital for maritime surveillance and safety.

Purpose of the Study:

• To develop an advanced fusion method for enhanced ship radiated noise recognition.
• To improve classification accuracy under low signal-to-noise ratio (SNR) conditions.
• To effectively integrate heterogeneous data and network features.

Main Methods:

• A machine learning-Dempster-Shafer (ML-DS) decision fusion method was proposed.
• Deep learning (CNN, LSTM) classified spectrogram and amplitude features.
• Machine learning (SVM) classified chromaticity features.
• A basic probability assignment model (BPA) fused classifier outputs using Dempster-Shafer theory.

Main Results:

• The ML-DS method significantly improved recognition for low SNR datasets.
• Lowest fusion recognition reached 76.01%, with an average of 94.92%.
• Outperformed traditional single-feature and one-step fusion algorithms.

Conclusions:

• The ML-DS method effectively fuses heterogeneous data and networks for robust identification.
• This approach offers a substantial improvement in ship radiated noise recognition accuracy.
• The ML-DS method is applicable to real-world ship radiated noise identification challenges.