Feature Extraction and Classification Recognition of Ship Radiated Noise Based on 2D-ACVMD

Due to the complex Marine environment and many noise components, it is very challenging to classify ship radiated noise. A time-frequency feature extraction and recognition method based on 2D-ACVMD and the Transformer network is proposed in this paper. After the ship radiated noise signal is preprocessed, it is analyzed by AOK time-frequency, and then the time-frequency graph is decomposed by 2D-ACVMD. Finally, the texture features of gray co-occurrence matrix are extracted from the decomposed time-frequency graph. Transformer network is selected as a classifier for classification recognition. Based on the ShipsEar dataset, with classification accuracy up to 98.8%, the experimental results show that the proposed method is feasible and advanced.