Underwater Acoustic Target Recognition Based on Supervised Feature-Separation Algorithm.

Area of Science:

• Acoustics
• Signal Processing
• Machine Learning

Background:

• Underwater acoustic target recognition is crucial but challenging with limited labeled data.
• Existing feature extraction methods struggle with small datasets.

Purpose of the Study:

• To develop a robust underwater acoustic target recognition method.
• To improve recognition accuracy using limited labeled data.

Main Methods:

• Resonance-based Sparsity Signal Decomposition (RSSD) for feature extraction.
• Unsupervised pre-training of a 1D convolution autoencoder-decoder for feature learning.
• Supervised fine-tuning for feature separation and classifier training.

Main Results:

• The proposed method achieved a recognition accuracy of 93.28% on ship-radiated noise data.
• The approach effectively utilizes unlabeled data for pre-training, reducing reliance on labeled data.
• Demonstrated superior performance compared to traditional state-of-the-art feature extraction techniques.

Conclusions:

• The novel 4-step method significantly enhances underwater acoustic target recognition accuracy.
• This technique is highly effective in scenarios with scarce labeled data.
• The combination of unsupervised pre-training and supervised fine-tuning offers a powerful solution for acoustic recognition challenges.