基于IVMD-RPE的船舶噪声信号去噪及特征提取方法.pdfVIP

基于IVMD-RPE的船舶噪声信号去噪及

特征提取方法

摘要：针对由于复杂环境条件下的船舶噪声信号识别度低的问题，提出了一种改进的变分模式分解(ImprovedVaria‐

tionalModeDecomposition,IVMD)、反向排列熵(ReversePermutationEntropy,RPE)、加权排列熵(WeightedPermuta‐

tionEntropy,WPE)和能量比相结合的船舶噪声信号识别分类方法。该方法利用IVMD将信号分解为若干本征模态函

数(IntrinsicModeFunction,IMF)；再利用RPE对IMF进行筛选，得到敏感IMF，实现去噪过程；最后计算IMF的

WPE并与各个IMF的能量比构建特征向量，建立长短期记忆网络(LongShort-TermMemory，LSTM)进行识别分类。

实验结果表明，与经验模态分解相比，文中提出的基于IVMD-RPE的船舶噪声信号去噪及特征提取方法能有效减少

环境噪声的影响，提高信噪比，对船舶噪声目标信号识别分类的准确率更高。

关键词：变分模式分解(VMD)；反向排列熵(RPE)；长短期记忆网络(LSTM)

Abstract:Aimingatthepracticalproblemoflowrecognitiondegreeofship-radiatednoiseundercomplex

environmentalconditions,amethodcombiningimprovedvariationalmodedecomposition(IVMD),reverse

permutationentropy(RPE),weightedpermutationentropy(WPE)andenergyratioisproposedforshipnoisesignal

recognitionandclassification.Inthismethod,IVMDisusedtodecomposethesignalintoseveralintrinsicmode

functions(IMFs),RPEisusedtoscreenIMFstoobtainsensitiveIMFsandrealizethedenoisingprocess,andWPEof

IMFiscalculatedandcomparedwiththeenergyratioofeachIMF.Finally,thefeaturevectorisconstructedandthe

longshort-termmemory(LSTM)networkisestablishedforrecognitionandclassification.Theexperimentalresults

showthat,comparedwiththeempiricalmodedecomposition,theproposedmethodbasedonIVMDandRPEcan

effectivelyreducetheinfluenceofenvironmentalnoise,improvethesignal-to-noiseratio,andhaveahigheraccuracyin

theidentificationandclassificationofshipnoisesignals.

Keywords:variationalmodedecomposition(VMD);reversepermutationentropy(RPE);longshorttermmemory(LSTM)