Extraction of noise feature and classification of underwater targets based on variational mode decomposition algorithm

When there are intermittent phenomena caused by abnormal events in the signal, the traditional empirical mode decomposition (EMD) algorithms often produce more severe mode aliasing, which significantly affects the performance of target feature extraction. In this paper, the variational mode decomposition (VMD) algorithm is used in the feature analysis and extraction of underwater target signals. This method can adaptively cut the signal frequency band, which largely avoids the mode aliasing phenomenon produced by the traditional EMD algorithm, and improves the accuracy of target feature extraction. At the same time, invalid calculations are avoided. In addition, the correlation threshold is used to select the modes to eliminate the interference modes to a certain extent. Based on the Hilbert transform of the mode function of each order of VMD, a VMD-HT feature set is proposed for target classification. Four classifiers are used to classify and recognize three kinds of underwater target noise signals. The comparison of classification results show that the VMD-HT feature set has better classification performance than other mode decomposition algorithms.