Feature extraction for sound source material recognition with impact sounds

    在线阅读 下载PDF 导出详情
    摘要 Soundsourcerecognitionisapartofenvironmentalsoundrecognition,whichisoneofthemostimportantresearchareasinpatternrecognition.Impactsoundscarrymuchphysicalinformationassociatedwiththesoundsources,whichmakesimpactsoundbasedsoundsourcerecognitionanimportantapproachtoimproverecognitionperformance.Inthisstudy,theimpactsoundcontinuumsynthesizedwithaball-platecollisionmodelisusedformaterialrecognitionoftheimpactedplates.Thebasisfunctionlearningmethodandtime-frequencyrepresentationmethods,includingtheshorttimeFouriertransformandthewaveletpackettransform,areappliedintoclassificationandtherecognitionresultsarecompared.TheresultshowsthatthefeaturesobtainedbyusingthebasisfunctionlearningperformbetterformaterialclassificationoftheimpactedplatesthanthatbyusingtheshorttimeFouriertransformandthewaveletpackettransform.Thisdemonstratesthehighefficiencyandsuperiorityofthismethodinmaterialrecognitionofsoundsources.
    机构地区 不详
    出处 《声学学报:英文版》 2017年2期
    关键词
    出版日期 2017年02月12日(中国Betway体育网页登陆平台首次上网日期,不代表论文的发表时间)
    • 相关文献
    Baidu
    map