Table Of Content

David Zhang · Kebin Wu Pathological Voice Analysis Pathological Voice Analysis (cid:129) David Zhang Kebin Wu Pathological Voice Analysis DavidZhang KebinWu TheChineseUniversity HuaweiTechnologies ofHongKong(Shenzhen) Beijing,China ShenzhenInstituteofArtificial IntelligenceandRoboticsforSociety Guangdong,China ISBN978-981-32-9195-9 ISBN978-981-32-9196-6 (eBook) https://doi.org/10.1007/978-981-32-9196-6 ©SpringerNatureSingaporePteLtd.2020 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartofthe materialisconcerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation, broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionorinformation storageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilarmethodology nowknownorhereafterdeveloped. Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse. The publisher, the authors, and the editorsare safeto assume that the adviceand informationin this bookarebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsor theeditorsgiveawarranty,expressedorimplied,withrespecttothematerialcontainedhereinorforany errorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictional claimsinpublishedmapsandinstitutionalaffiliations. ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSingaporePteLtd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore Preface The human voice contains rich information about the speaker, such as identity, gender, health, situation, and emotion. However, the biomedical value of voice is lessaddressedcomparedwithitsbiometricsapplicationssuchasspeechrecognition and speaker identification. In this book, we present the importance and value of voiceinaidingdiagnosisthroughpathologicalvoiceanalysis. We systematically introduce our research works on pathological voice analysis fromthefollowingthreeaspects:(1)areviewonpathologicalvoiceanalysisanda guidelineonvoiceacquisitionforclinicalapplication;(2)designappropriatesignal processingalgorithmsforpathologicalvoice;and(3)extractbiomedicalinformation invoicetoimprovediseasedetection,suchasdictionary-basedfeaturelearningand multi-audio fusion. Experimental results have shown the superiority of these tech- niques. These proposed methods can be used in applications of voice quality assessment, disease detection, disease severity prediction, and even the analysis of other similar signals. This book will be useful to researchers, professionals, and postgraduate students working in the field of speech signal processing, pattern recognition, biomedical engineering, etc. This book also will be very meaningful forinterdisciplinaryresearch. The book is organized as follows: In Chap. 1, the development of pathological voiceanalysisissystematicallyreviewed.Thenimportantfactorsintheacquisition of pathological voice, sampling rate in particular, are discussed in Chap. 2. After that, two of the widely used signal processing steps in pathological voice analysis, whicharepitchestimationandglottalclosureinstant(GCI)detection,arestudiedin Chaps. 3 and 4, respectively. In Chap. 5, feature learning based on spherical K-means is proposed, in contrast to the traditional handcrafted features. In Chaps.6and7,weinvestigatemulti-audiofusioninpathologicalvoiceanalysisso as to make full use of the multiple audios collected for each subject. Finally, we summarizethebookandgiveashortintroductiontofutureworksthatmaycontrib- utetothefurtherdevelopmentofpathologicalvoiceanalysisinChap.8. Ourteamhasbeenworkingonpathologicalvoiceanalysisformorethan5years. We appreciate the related grant supports from the GRF fund of the HKSAR v vi Preface Government,ResearchprojectsfromShenzhenInstitutesofbothBigDataandArti- ficial Intelligence & Robotics for Society, and the National Natural Science Foun- dation of China (NSFC) (61020106004, 61332011, 61272292, and 61271344). Besides, we thank Prof. Guangming Lu and Prof. Zhenhua Guo for their valuable suggestionsaboutourresearchonpathologicalvoiceanalysis. Guangdong,China DavidZhang Beijing,China KebinWu April2020 Contents 1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 PathologicalVoiceAnalysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 ComputerizedVoiceAnalysis. . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2.2 BiomedicalValueofVoice. . . . . . . . . . . . . . . . . . . . . . . 5 1.2.3 PresentSituationofComputerizedVoiceAnalysis. . . . . . . 11 1.2.4 OpenChallengesinComputerizedVoiceAnalysis. . . . . . . 17 1.2.5 FutureDirectionsinComputerizedVoiceAnalysis. . . . . . 20 1.2.6 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2 PathologicalVoiceAcquisition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 2.2 DatabaseDescription. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 2.3 ExperimentalResults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.3.1 Metric1:InformationEntropy(SignalLevel). . . . . . . . . . 35 2.3.2 Metric2:ReconstructionError(SignalLevel). . . . . . . . . . 36 2.3.3 Metric3:FeatureCorrelation(FeatureLevel). . . . . . . . . . 37 2.3.4 Metric4:ClassificationAccuracy(SystemLevel). . . . . . . 38 2.3.5 Metric5:ComputationalCost(HardwareLevel). . . . . . . . 39 2.3.6 Metric6:StorageCost(HardwareLevel). . . .. . . .. . . . .. 40 2.3.7 DiscussionandGuidelineontheSamplingRate Selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2.4 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 3 PitchEstimation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.2 RelatedWorks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 vii viii Contents 3.3 HarmonicsEnhancementforPitchEstimation. . . . . . . . . . . .. . . . 52 3.3.1 MotivationforHarmonicsEnhancement. . . . . . . . . . . . . . 52 3.3.2 TheoreticalAnalysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 3.3.3 Implementation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 3.3.4 In-DepthAnalysisoftheProposedAlgorithm. . . . . . . . . . 60 3.4 ExperimentalResult. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.4.1 BasicExperimentalSetting. . . . . . . . . . . . . . . . . . . . . . . . 65 3.4.2 PerformancefortheNoisySpeeches. . . . . . . . . . . . . . . . . 66 3.4.3 ExtensiontoMusic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 3.4.4 Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 3.5 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4 GlottalClosureInstantsDetection. . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 4.2 RelatedWorks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 4.2.1 ClassicalGCIDetectionMethods. . . . . . . . . . . . . . . . . . . 79 4.2.2 Limitations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 4.3 GCIDetectionUsingTKEO. . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 4.3.1 TKEOandItsRelationshipwithGCI. . . . . . . . . . . . . . . . 82 4.3.2 AbsoluteTKEOwithScaleParameterk. . . . . . . . . . . . . . 83 4.3.3 MultiresolutionCombination. . . . . . . . . . . . . . . . . . . . . . 84 4.3.4 GMAT:GCIDetectionBasedonMultiresolution AbsoluteTKEO. .. . . . . . . . . . . . . .. . . . . . . . . . . . . .. . 86 4.4 ExperimentalResults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 4.4.1 BasicExperimentalSettings. . . . . . . . . . . . . . . . . . . . . . . 88 4.4.2 PerformanceComparison. . . . . . . . . . . . . . . . . . . . . . . . . 90 4.4.3 GCIsDetectioninPathologicalSpeechIdentification. . . . . 97 4.5 Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 4.5.1 ParameterSensitivity. . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 4.5.2 GMATandLowFrequencyNoise. . . . . . . . . . . . . . . . . . 101 4.5.3 RelationBetweenGMATandMSM. . . . . . . . . . . . . . . . . 101 4.5.4 DifferentPoolingMethodsinGMAT. . . . . . . . . . . . . . . . 102 4.6 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 5 FeatureLearning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5.2 RelatedWorks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 5.3 ProposedFeatureLearningMethod. . . . . . . . . . . . . . . . . . . . . . . 110 5.3.1 Preprocessing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 5.3.2 Mel-Spectrogram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 5.3.3 DictionaryLearningBasedonSphericalK-Means. . . . . . . 112 5.3.4 EncodingandPooling. . . . . . . . . . . . . . . . . . . . . . . . . . . 112 Contents ix 5.4 ExperimentalResults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 5.4.1 DatasetandExperimentalSetup. . . . . . . . . . . . . . . . . . . . 113 5.4.2 PDDetectionwithLearnedandHand-CraftedFeatures. . . 114 5.4.3 AblationStudy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 5.5 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 6 JointLearningforVoiceBasedDiseaseDetection. . . . . . . . . . . . . . . 123 6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 6.2 RelatedWorks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 6.2.1 Notation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 6.2.2 RidgeRegression. . . .. . . . . . . . . . . . .. . . . . . . . . . . .. . 125 6.2.3 Low-RankRidgeRegression. . . . . . . . . . . . . . . . . . . . . . 126 6.2.4 ε‐DraggingTechnique. . . . . . . . . . . . . . . . . . . . . . . . . . . 127 6.3 ProposedMethod. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 6.3.1 JointLearningwithLabelRelaxedLow-RankRidge Regression(JOLL4R). . . . . . . . . . . . . . . . . . . . . . . . . . . 128 6.3.2 AlgorithmtoSolveJOLL4R. .. . . . . . .. . . . . .. . . . . . .. 129 6.3.3 Classification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 6.4 ExperimentalResults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 6.4.1 DatasetandExperimentalSetup. . . . . . . . . . . . . . . . . . . . 134 6.4.2 TheDetectionofPatientswithCordectomy. . . . . . . . . . . . 135 6.4.3 TheDetectionofPatientswithFrontolateralResection. . . . 135 6.4.4 AblationStudy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 6.5 Discussions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 6.5.1 VisualizationofJOLL4R. . . . . . . . . . . . . . . . . . . . . . . . . 139 6.5.2 ClassifierinJOLL4R. . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 6.5.3 ExtensiontoFusionofMoreAudios. . . . . . . . . . . . . . . . . 141 6.6 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 7 RobustMulti-ViewDiscriminativeLearningforVoiceBased DiseaseDetection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 7.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 7.2 ProposedMethod. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 7.2.1 Notation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 7.2.2 TheObjectiveFunction. . . . . . . . . . . . . . . . . . . . . . . . . . 149 7.2.3 OptimizationofROME-DLR. . . . . . . . . . . . . . . . . . . . . . 152 7.2.4 TheClassificationRuleofROME-DLR. . . . . . . . . . . . . . 155 7.3 ExperimentalResults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 7.3.1 DatasetandExperimentalSetup. . . . . . . . . . . . . . . . . . . . 156 7.3.2 TheDetectionofRLNP. . . . . . . . . . . . . . . . . . . . . . . . . . 157 7.3.3 TheDetectionofSD. . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 7.3.4 AblationStudy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 x Contents 7.4 Discussions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 7.5 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 8 BookReviewandFutureWork. . . . . . .. . . . . . . . . .. . . . . . . . . . .. 167 8.1 BookRecapitulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 8.2 FutureWork. . . .. . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . 169 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

ebook img

Pathological Voice Analysis PDF

181 Pages·2020·4.648 MB·English
by  David Zhang, Kebin Wu
#
Save to my drive
Quick download
Download

Download Pathological Voice Analysis PDF Free - Full Version

by David Zhang, Kebin Wu| 2020| 181 pages| 4.648| English

Download Pathological Voice Analysis by David Zhang, Kebin Wu in PDF format completely FREE. No registration required, no payment needed. Get instant access to this valuable resource on PDFdrive.to!

Free Download PDF

About Pathological Voice Analysis

No description available for this book.

Detailed Information

Author:David Zhang, Kebin Wu
Publication Year:2020
ISBN:9789813291966
Pages:181
Language:English
File Size:4.648
Format:PDF
Price:FREE
Download Free PDF

Safe & Secure Download - No registration required

Why Choose PDFdrive for Your Free Pathological Voice Analysis Download?

  • 100% Free: No hidden fees or subscriptions required for one book every day.
  • No Registration: Immediate access is available without creating accounts for one book every day.
  • Safe and Secure: Clean downloads without malware or viruses
  • Multiple Formats: PDF, MOBI, Mpub,... optimized for all devices
  • Educational Resource: Supporting knowledge sharing and learning

Free Books Similar to Pathological Voice Analysis

Frequently Asked Questions

Is it really free to download Pathological Voice Analysis PDF?

Yes, on https://PDFdrive.to you can download Pathological Voice Analysis by David Zhang, Kebin Wu completely free. We don't require any payment, subscription, or registration to access this PDF file. For 3 books every day.

How can I read Pathological Voice Analysis on my mobile device?

After downloading Pathological Voice Analysis PDF, you can open it with any PDF reader app on your phone or tablet. We recommend using Adobe Acrobat Reader, Apple Books, or Google Play Books for the best reading experience.

Is this the full version of Pathological Voice Analysis?

Yes, this is the complete PDF version of Pathological Voice Analysis by David Zhang, Kebin Wu. You will be able to read the entire content as in the printed version without missing any pages.

Is it legal to download Pathological Voice Analysis PDF for free?

https://PDFdrive.to provides links to free educational resources available online. We do not store any files on our servers. Please be aware of copyright laws in your country before downloading.

The materials shared are intended for research, educational, and personal use in accordance with fair use principles.

We use cookies

We use cookies to ensure you get the best browsing experience on our website. By clicking "Accept Cookies", you agree that we can store cookies on your device in accordance with our Terms and Privacy.