Heart Sound Segmentation Based on Personalized Gaussian Mixture Model and Convolutional Neural Network

Heart Sound Segmentation Based on Personalized Gaussian Mixture Model and Convolutional Neural Network | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research article Heart Sound Segmentation Based on Personalized Gaussian Mixture Model and Convolutional Neural Network Chun Dong Xu, Jing Zhou, Dong Wen Ying, Lei Jing Hou, Qing Hua Long This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.2.20414/v2 This work is licensed under a CC BY 4.0 License Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Abstract Background: Heart sound segmentation is a long-standing problem in heart analysis, and it is mainly caused by noise interference and diversification of heart sounds. Faced with the challenging of heart sound segmentation, a more applicable segmentation model was studied. Methods: In this process, the optimal modified Log-spectral amplitude and wavelet were used to suppress the noise in the heart sound, and used the duration-dependent hidden Markov model based on personalized Gaussian mixture model (PGMM-DHMM) to segment the fundamental heart sound (FHS) and the non-fundamental heart sound (non-FHS). Then used the optimized Mel frequency cepstral coefficient (MFCC) to realize the classification of S1 and S2 heart sound frames through the Convolutional neural network (CNN) classifier, which can avoid the errors caused by the ambiguity of the time domain features. Results: PGMM-DHMM can segment FHS more effectively, and the accuracy is 94.3%. The CNN classifier obtained the best results in the S1 and S2 classifications, the accuracy is 90.92%, the precision of S1 is 90.76%, the recall is 91.05%, the F-measure is 90.9%, and the precision of S2 is 91.07%, the recall is 90.79%, the F-measure is 90.93%. The final segmentation accuracy is 92.92%. In addition, the experimental results further indicate that CNN has more robust performance when classifying abnormal S1 and abnormal S2. Conclusions: The PGMM-DHMM model can better segment FHS and Non-FHS. The optimization of MFCC improves the classification effect of S1 and S2, and the improvement effect by the CNN classifier is significant, especially for abnormal heart sounds. The proposed algorithm is better than other algorithms at this stage. Medical Informatics Mel frequency cepstral coefficient heart sound segmentation heart sound classification Gaussian mixture model Convolutional neural network Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Full Text Cite Share Download PDF Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-11021","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research article","associatedPublications":[],"authors":[{"id":3510248,"identity":"fc31698a-7e0c-408d-97d3-ab95c4bc6057","order_by":0,"name":"Chun Dong Xu","email":"","orcid":"","institution":"Jiangxi University of Science and Technology","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Chun","middleName":"Dong","lastName":"Xu","suffix":""},{"id":3510249,"identity":"0326295c-d13f-4a02-9d2c-3876d60a5073","order_by":1,"name":"Jing Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzklEQVRIiWNgGAWjYBACNobzDx9+MJCQs+9vPkCcFj7GM8zGEgUWxgYSxxKI0yLHfIZNgudDReIGhhwDIh3GdvawgYSBhLE5w5mPN94w2MnpNhDSwnMu8UEB0C+Wzb2bLecwJBubHSCkReKAMdgWhgNnt0nzMBxI3EZQi/wDMwkeA4nEhgM5z4jUwnAGomXDgRw2YrUcSzYGOUxyxjFjyzkGRPhFvuHwwYcf/tTJ8fM3P7zxpsJOjqAWFAB0ISnKIVpI1TEKRsEoGAUjAgAAc2NB9zgooMIAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-7945-3680","institution":"Jiangxi University of Technology","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Jing","middleName":"","lastName":"Zhou","suffix":""},{"id":3510250,"identity":"fe10c7fd-76d4-499a-9e3a-5789fd543822","order_by":2,"name":"Dong Wen Ying","email":"","orcid":"","institution":"University of the Chinese Academy of Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Dong","middleName":"Wen","lastName":"Ying","suffix":""},{"id":3510251,"identity":"03e6032c-5bb8-4b09-8fb2-44e055625c77","order_by":3,"name":"Lei Jing Hou","email":"","orcid":"","institution":"key Laboratory of Speech Acoustics and Content Understanding, Institute of Acoustics","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Lei","middleName":"Jing","lastName":"Hou","suffix":""},{"id":3510252,"identity":"226b9c68-9a48-42f2-aeff-027f155ea6bb","order_by":4,"name":"Qing Hua Long","email":"","orcid":"","institution":"Jiangxi University of Sciences and Technology","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Qing","middleName":"Hua","lastName":"Long","suffix":""}],"badges":[],"createdAt":"2020-01-07 13:14:21","currentVersionCode":2,"declarations":"","doi":"10.21203/rs.2.20414/v2","doiUrl":"https://doi.org/10.21203/rs.2.20414/v2","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":3139690,"identity":"cff6fa03-b60d-4745-a3c2-23f907370667","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":84675,"visible":true,"origin":"","legend":"Detailed flow chart of the proposed algorithm","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/3095605625c7858ac19cdd06.png"},{"id":3139691,"identity":"988ed410-4d61-4b10-972b-e6ce46109339","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":95174,"visible":true,"origin":"","legend":"Test results of noise reduction by OMLSA-W. (a1) The original a0001, (a2) The denoised a0001, (b1) The original b0087, (b2) The denoised b0087, (c1) The original c0018, (c2) The denoised c0018, (d1) The original d0052, (d2) The denoised d0052.","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/0eac2548969fe1aed2f27ba4.png"},{"id":3139692,"identity":"9a72c8fc-7578-4b9d-954b-d4b07487942b","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":70537,"visible":true,"origin":"","legend":"State transition between FHS and non-FHS","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/d791f70fa7c6ccb1eda6c10b.png"},{"id":3139693,"identity":"bb1fac9f-3fa1-4980-9000-790ba80dc0c3","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":61868,"visible":true,"origin":"","legend":"Detection of cardiac cycle","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/7bbf34b446c1fe01b4d2d110.png"},{"id":3139694,"identity":"97765670-3296-4525-8be7-0b1f1eecce45","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":44194,"visible":true,"origin":"","legend":"Gaussian modeling of eigenmode cardiac cycle","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/540e62f3313819a1a04aa09a.png"},{"id":3139695,"identity":"a9474acc-efde-4374-9f3d-394623060eb5","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":63045,"visible":true,"origin":"","legend":"The cardiac cycle segmentation results of a0042","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/2e043ae6192e7d88423036a0.png"},{"id":3139696,"identity":"fda45ac5-8c5a-49be-b390-6c122b2985ca","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":192571,"visible":true,"origin":"","legend":"Visualization of MFCC: (a) Each dimension of MFCC, (b) 3D map of MFCC.","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/215172edba74858ea134ef12.png"},{"id":3139697,"identity":"30b0a36c-ff87-42c8-9c5d-1803ae2eefe2","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":62038,"visible":true,"origin":"","legend":"The test results of frequency scale conversion parameter correction","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/ef6a315594c5ef5ae9108002.png"},{"id":3139698,"identity":"41ab608e-5005-4524-ba8f-b1177af6e85b","added_by":"auto","created_at":"2020-10-22 20:01:49","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":54171,"visible":true,"origin":"","legend":"The structure of DNN classifier","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/2dc88050625adbe2926e1619.png"},{"id":3139699,"identity":"00acf3b2-3916-4cce-82de-1c8b33424ae9","added_by":"auto","created_at":"2020-10-22 20:01:50","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":53134,"visible":true,"origin":"","legend":"CNN classifier","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/504926bcdf4da38bca5ff9c0.png"},{"id":3139700,"identity":"e9da3e9c-2d5d-4092-9471-78f2316a97f2","added_by":"auto","created_at":"2020-10-22 20:01:50","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":108136,"visible":true,"origin":"","legend":"The training results, (a) Optimizing MFCC testing, (b) Fusion feature testing, (c) Classifier testing.","description":"","filename":"11.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/875569811d6362703e453e5c.png"},{"id":3139701,"identity":"6286c4df-350b-4676-9da0-3bb21be01357","added_by":"auto","created_at":"2020-10-22 20:01:50","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":130046,"visible":true,"origin":"","legend":"(a) The evaluation indexes of abnormality S1, (b) The evaluation indexes of the normal S1, (c) The evaluation indexes of the abnormality S2, (d) the evaluation indexes of the normal S2","description":"","filename":"12.png","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2/50d8e8c3990fa1200a587886.png"},{"id":13545772,"identity":"ff113d5c-c846-41e4-81df-4a1f3d25026a","added_by":"auto","created_at":"2021-09-17 02:07:25","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2060567,"visible":true,"origin":"","legend":"","description":"","filename":"HEARTSOUNDSEGMENTATIONRevisedManuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2_covered.pdf"},{"id":13484865,"identity":"c5ff6d2c-c636-4f67-8d93-c7a2abd29885","added_by":"auto","created_at":"2021-09-16 21:59:39","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2930600,"visible":true,"origin":"","legend":"","description":"","filename":"HEARTSOUNDSEGMENTATIONRevisedManuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-11021/v1_covered.pdf"},{"id":3139706,"identity":"7feec13d-34ee-484c-8252-3c390f085609","added_by":"auto","created_at":"2020-10-22 20:02:19","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2362560,"visible":true,"origin":"","legend":"","description":"","filename":"HEARTSOUNDSEGMENTATIONRevisedManuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-11021/v2_stamped.pdf"}],"financialInterests":"","formattedTitle":"Heart Sound Segmentation Based on Personalized Gaussian Mixture Model and Convolutional Neural Network","fulltext":[{"header":"Full Text","content":"\u003cp\u003eThis preprint is available for \u003ca href='/article/rs-11021/latest.pdf' target='_blank'\u003edownload as a PDF\u003c/a\u003e.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Mel frequency cepstral coefficient, heart sound segmentation, heart sound classification, Gaussian mixture model, Convolutional neural network","lastPublishedDoi":"10.21203/rs.2.20414/v2","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.2.20414/v2","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground:\u003c/strong\u003e Heart sound segmentation is a long-standing problem in heart analysis, and it is mainly caused by noise interference and diversification of heart sounds. Faced with the challenging of heart sound segmentation, a more applicable segmentation model was studied. \u003c/p\u003e\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e In this process, the optimal modified Log-spectral amplitude and wavelet were used to suppress the noise in the heart sound, and used the duration-dependent hidden Markov model based on personalized Gaussian mixture model (PGMM-DHMM) to segment the fundamental heart sound (FHS) and the non-fundamental heart sound (non-FHS). Then used the optimized Mel frequency cepstral coefficient (MFCC) to realize the classification of S1 and S2 heart sound frames through the Convolutional neural network (CNN) classifier, which can avoid the errors caused by the ambiguity of the time domain features. \u003c/p\u003e\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e PGMM-DHMM can segment FHS more effectively, and the accuracy is 94.3%. The CNN classifier obtained the best results in the S1 and S2 classifications, the accuracy is 90.92%, the precision of S1 is 90.76%, the recall is 91.05%, the F-measure is 90.9%, and the precision of S2 is 91.07%, the recall is 90.79%, the F-measure is 90.93%. The final segmentation accuracy is 92.92%. In addition, the experimental results further indicate that CNN has more robust performance when classifying abnormal S1 and abnormal S2. \u003c/p\u003e\u003cp\u003e\u003cstrong\u003eConclusions:\u003c/strong\u003e The PGMM-DHMM model can better segment FHS and Non-FHS. The optimization of MFCC improves the classification effect of S1 and S2, and the improvement effect by the CNN classifier is significant, especially for abnormal heart sounds. The proposed algorithm is better than other algorithms at this stage.\u003c/p\u003e","manuscriptTitle":"Heart Sound Segmentation Based on Personalized Gaussian Mixture Model and Convolutional Neural Network","msid":"","msnumber":"","nonDraftVersions":[{"code":2,"date":"2020-10-22 20:01:44","doi":"10.21203/rs.2.20414/v2","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}},{"code":1,"date":"2020-01-09 19:09:37","doi":"10.21203/rs.2.20414/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0c141367-5911-43f1-ae0f-afaeaa4cd416","owner":[],"postedDate":"October 22nd, 2020","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":48362,"name":"Medical Informatics"}],"tags":[],"updatedAt":"2020-10-15T11:50:50+00:00","versionOfRecord":[],"versionCreatedAt":"2020-10-22 20:01:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v2","identity":"rs-11021","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"identity":"rs-11021","version":["v2"]},"buildId":"WrCJVZZCHTDjtuVLN7oU0","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}