Offshore ship classification based on Mel-frequency Cepstral Coefficients of main intrinsic mode

Recognition of ship-radiated noise is essential for advancing ship automation, marine resource exploration, and underwater operations. This study presents a novel recognition method for ship signals that integrates the variational mode decomposition (VMD) with Mel-frequency cepstral coefficients (MFCC). The method decomposes ship-radiated noise into the intrinsic mode functions (IMFs) and selects the top IMF that predominantly contributes to the signal's energy, which is then used to extract the MFCC of the ship-radiated noise signals. The extracted MFCC features are used to recognize different ships with the support vector machine (SVM). The experimental results show that this method achieves an average recognition rate exceeding 98% for the three types of ship-radiated noises across ShipsEar database. The proposed VMD-MFCC-SVM (VMS) not only enhances the recognition of ship signals, but also offers a promising approach for the identification of ship signals, highlighting its substantial potential for practical applications.