水声目标识别中的多尺度特征提取方法研究

打开文本图片集

中图分类号：TN911.7；TB566 文献标识码：A文章编号：2096-4706（2025）16-0008-07

Research on Multi-scale Feature Extraction Methods for Underwater Acoustic Target Recognition

SUN Dan

（InnovationInstituteofEngineeringEducationandEngineeringCulture，Hunan InstituteofEngineering，Xiangtan 411104， China）

Abstract： Underwater acoustic target recognition is an important research topic in the field of underwater acoustic informationprocesing.Itsfeatureextractionandrecognitionfacechallengessuchascomplexunderwateracousticenvironment andvariabletargetfeatures.Inthispaper，theteoreticalbasis，algorithmimplementationandperformancecharactersticsoftwo multi-scale feature extraction methods basedon Mel-frequency Cepstral Coeffcients （MFCC）and Hilbert-Huang Transform （HHT）are analyzedin depth.Theresearch shows that the MFCCmethod realizes the nonlinear mapping through the auditory perception characteristics basedonthe Mel frequencyscaleand the short-time Fourier transform，and effectively expreses the spectral characteristicsof theunderwateracoustic signal.The HHTmethod is basedontheadaptive decompositionframework， whichcanaccuratelycapture the dynamiccharacteristicsofunderwateracoustic targets.Aimingat thespecialrequirements ofunderwater acoustic target recognition，this paper proposes optimization strategies such as parameter optimization based on time-frequencyresolution trade-offand componentscreening basedon energy-frequency distribution，and establishesa systematiccomparativeaalysis framework fromthreedimensionsofcomputationalcomplexityfeatureexpresionabilityand environmental adaptability， which provides a theoretical basis for the selection of methods in practical applications.

Keywords： underwater acoustic target recognition; feature extraction; MFCC; HHT; multi-scale analysis

0 引言

水声目标识别是水声信息处理领域的重要研究课题，对海洋资源开发、海域安全监测、水下目标探测等具有重要的现实意义。（剩余11071字）