2023年8月第19卷第3期系统仿真技术SystemSimulationTechnologyAug.,2023Vol.19,No.3基于快速傅里叶变换的乐器音色特征识别算法研究赵逸凡1,佟庆月1,卢艺1*,戴卫2,廖仲楠2(1.上海交通大学电子信息与电气工程学院自动化系,上海200240;2.上海戴氏琴弦制作社,上海201100)摘要:本研究主要以传统乐器为研究对象,探究了乐器的音色本质。在频域中利用快速傅里叶变换提出了乐音泛音列提取的2种方法,即最大倍频和法和多倍频法,并将二者在不同场景下进行了比较。发现多倍频法具有较好的鲁棒性。同时,考虑到乐音在时间维度上也蕴含一定的信息,进而针对性地使用语谱图以弥补传统频谱中一些信息丢失的问题,并根据提取到的泛音列支持了乐器音色特征识别,在中国传统乐器古琴中进行了算法测试,结果表明本研究所构建的方法可推广到其他中西乐器,未来还可以拓宽到信号特征辨识、材料探伤等领域。关键词:语谱图;音色特征;泛音列提取;快速傅里叶变换ResearchonMusicalInstrumentTimbreFeatureIdentificationAlgorithmBasedonFastFourierTransformZHAOYifan1,TONGQingyue1,LUYi1*,DAIWei2,LIAOZhongnan2(1.AutomationDepartment,SchoolofElectronicInformationandElectricalEngineering,ShanghaiJiaoTongUniversity,Shanghai200240,China;2.ShanghaiDai’sStringCompany,Shanghai201100,China)Abstract:Thetimbralnatureofmusicalinstrumentsisexplored,mainlyusingtraditionalmusicalinstrumentsastheresearchobject.TwomethodsofmusicalovertonecolumnextractionareproposedinthefrequencydomainusingthefastFouriertransform,andthetwoareexploredincomparisonindifferentscenarios.Themultifrequencymethodisfoundtohavebetterrobustness.Atthesametime,consideringthatmusicaltonesalsocontaincertaininformationinthetimedimension,andthenthetargeteduseofthespeechspectrogramtocompensateforsomeinformationlossinthetraditionalspectrum,andbasedontheextractedovertonecolumnstosupportthetimbrefeaturerecognitionofmusicalinstruments,thealgorithmwastestedonthetraditionalChineseinstrumentGuqin,showingthatthemethodconstructedinthisstudycanbeextendedtootherChineseandWesternmusicalinstruments,andcanalsobebroadenedtosignalfeaturerecognition,materialdetectionandotherfieldsinthefuture,mat...