基于变分模态分解的水下目标噪声特征提取及分类

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

当信号中存在异常事件引起的间歇现象时,传统的经验模态分解算法常易产生较为严重的模态混叠现象,严重影响目标特征提取的性能。文章在水下被动目标信号特征分析提取中引用变分模态分解算法。该方法能够自适应地对信号频带进行切割,极大程度上避免了传统模态分解算法所产生的模态混叠现象,提高了对目标特征提取的准确性,同时也避免了无效计算。此外,还利用相关性阈值进行模态选择,一定程度上消除干扰模态。在对变分模态分解(Variational Mode Decomposition, VMD)的各阶模态函数进行希尔伯特变换的基础上,提出一种基于变分模态分解和希尔伯特变换(VDM-Hilbert Transformation, VDM-HT)联合处理的特征集进行目标分类。采用四种分类器对3种水下目标噪声信号进行分类。结果表明,VMD-HT算法所提取的特征集相比其他模态分解算法具有更好的分类性能。

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.