Abstract: Aiming at the practical problem of low recognition degree of ship-radiated noise under complex environmental conditions, a method combining improved variational mode decomposition (IVMD), reverse permutation entropy (RPE), weighted permutation entropy (WPE) and energy ratio is proposed for ship noise signal recognition and classification. In this method, IVMD is used to decompose the signal into several intrinsic mode functions (IMFs), RPE is used to screen IMFs to obtain sensitive IMFs and realize the denoising process, and WPE of IMF is calculated and compared with the energy ratio of each IMF. Finally, the feature vector is constructed and the long short-term memory (LSTM) network is established for recognition and classification. The experimental results show that, compared with the empirical mode decomposition, the proposed method based on IVMD and RPE can effectively reduce the influence of environmental noise, improve the signal-to-noise ratio, and have a higher accuracy in the identification and classification of ship noise signals.