Ship Noise Detection Based on Light Weight Deep Neural Networks

Ship noise detection plays a vital role in improving maritime safety, protecting marine ecosystems, and improving vessel efficiency. Deep learning approaches have shown significant success in signal detection and classification tasks, leading to their adoption in real-world applications. Moreover, mobile computing demands applications with reduced storage size, low processing and memory requirements, and energy efficiency. In this paper, we propose a ship detection scheme based on light weight neural networks. By incorporating the Demodulation Envelope Detection (DEMON) spectral analysis method to feature extraction, the light weight models, MobileNet1D and AM- MobileNet1D, are designed to detect the ship. By classifying and processing for the acquired ship signal dataset, the experimental results demonstrate that the proposed scheme outperforms the benchmark algorithms. The proposed scheme in this paper can provide theoretical support for the hardware implementation and application of real-time ship noise detection.