SO-slope entropy coupled with SVMD: A novel adaptive feature extraction method for ship-radiated noise

Abstract

Slope entropy (SloEn) has been applied as a powerful nonlinear dynamic tool for signal complexity measurement and is widely used for ship-radiated noise signal (S-RNS) feature extraction. However, the thresholds of SloEn affect the entropy value, which influences the effect of feature extraction. Aimed at addressing the problem, this paper uses the snake optimizer (SO) to optimize SloEn and proposes a new entropy indicator, named SO SloEn (SO-SloEn), and then a novel adaptive S-RNS feature extraction method is put forward in combination with the successive variational mode decomposition (SVMD) and SO-SloEn, which solves the parameter selection problem of variational mode decomposition (VMD). First, SVMD is employed to decompose the S-RNS into several intrinsic mode functions (IMFs); after that, the SO-SloEn of IMFs is calculated to obtain the feature matrix dataset; finally, optimal IMF combinations are obtained through feature selection. The effectiveness of the proposed method is verified by two S-RNS cases, and the results indicate that the recognition rate of the proposed method is always the highest compared with other decomposition algorithms and other entropy indicators under the condition of extracting different numbers of IMFs; moreover, the highest recognition rate can reach more than 92% in both cases.