2023年第47卷第4期72语音技术语音技术oiceTechnologyV文献引用格式:陈锦辉.基于频域自适应动态压缩器的齿音抑制算法[J].电声技术,2023,47(4):72-74,85.CHENJH.Toothsoundsuppressionalgorithmbasedonfrequencydomainadaptivedynamiccompressor[J].AudioEngineering,2023,47(4):72-74,85.中图分类号:TP393文献标识码:ADOI:10.16311/j.audioe.2023.04.020基于频域自适应动态压缩器的齿音抑制算法陈锦辉(福建星网智慧科技有限公司,福建福州530002)摘要:在免提通话系统中,当语音中含有“z,c,s,zh,ch,sh”等嘶嘶作响的声母时,气流在唇齿间摩擦产生齿音,这种刺耳的齿音会分散听者的注意力并导致听觉疲劳,最终影响通话体验。为了解决这一问题,基于动态压缩器提出了一种基于频域自适应压缩器的齿音抑制算法。该算法将语音信号从时域转换到频域,并将其划分为10个子频带。每个子频带分别经过自适应白化和动态压缩处理,以减弱齿音所在频段的能量。实验结果表明,相比于均衡器,基于频域自适应动态压缩器的齿音抑制算法在齿音的识别和抑制强度方面表现更优,同时对无齿音的时段和频段没有抑制作用,降低了语音的响度衰减。关键词:齿音抑制;动态压缩器;自适应白化ToothSoundSuppressionAlgorithmBasedonFrequencyDomainAdaptiveDynamicCompressorCHENJinhui(FuJianI-StarTechnologyCo.,Ltd.,Fuzhou350002,China)Abstract:Inthehands-freecommunicationsystem,whenthevoicecontainshissinginitialssuchas“z,c,s,zh,ch,sh”,theairflowrubsbetweenthelipsandteeth,whichwilldistractthelistenerandleadtohearingfatigue,andultimatelyaffectthecommunicationexperience.Inordertosolvethisproblem,basedonthedynamiccompressor,atoothsoundsuppressionalgorithmbasedonthefrequencydomainadaptivecompressorisproposed.Thealgorithmtransformsthespeechsignalfromtimedomaintofrequencydomainanddividesitinto10sub-bands.Eachsub-bandissubjectedtoadaptivewhiteninganddynamiccompressiontoweakentheenergyofthefrequencybandwherethetoothsoundislocated.Theexperimentalresultsshowthat,comparedwiththeequalizer,thetoothsoundsuppressionalgorithmbasedonfrequencydomainadaptivedynamiccompressorperformsbetterintheidentificationandsuppressionintensityoftoothsound,andhasnosuppressioneffectontheperiodandfrequ...