A Reliable Bangla Sign Language Recognition System Using MediaPipe and LSTM Networks

A Reliable Bangla Sign Language Recognition System Using MediaPipe and LSTM Networks | 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 A Reliable Bangla Sign Language Recognition System Using MediaPipe and LSTM Networks Din Mohammad Toufik, Sumayae Binata Khaiat This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8949739/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Generally, the traditional mode of communication among Bangladesh's speech and hearing challenged patients is through the mode referred to as "Bangla Sign Language" (abbreviated as "BdSL"). Though the mode exists as an effective bridge to connect and communicate among patients and the general public, the lack of an appropriate interpretation mechanism has led to an enormous communication problem among patients and the public. In the present study, an efficient and "real-time" mode referred to as "Bangla Sign Language Recognition" (acronym "BSLR"), where "MediaPipe" and "Long Short-Term Memory" (acronym "LSTM" networks) are utilized, is proposed as follows: Instead of living up to the usual and more cumbersome "Convolutional Neural Network" (acronym "CNN" networks) based methods requiring image processing, skeleton "keypoints" have been considered for efficient "comprehension" and "perception". In order to ensure that the systems have adequate resistance from backgrounds and lighting changes, our team used holistics from Mediapipe to obtain relevant features (x, y, z coordinates) from both hands and body postures. To classify the gesture set, relevant features extracted from both hands and postures have been used as input to the specially developed LSTM scheme to classify the gesture set input to the specially developed LSTM scheme. In order to test and confirm our system, an entire dataset was prepared consisting of 50 BdSL signals (including alphabets and phrases) from 15 participants in 7,500 video clips. The proposed design has a training accuracy of 99.2% and a testing accuracy of 98.5%, as supported by the experimental results. Further, it has an inference latency as low as 40 ms, which allows it to be useful for the deployment of real-time communication systems. This work validates the assertion that for continuous Bangla Sign Language recognition, the combination of the effective features extracted by MediaPipe and the powerful sequence learning capabilities offered by LSTMs far surpass the state-of-the-art models based on static images. Cell Communication and Signaling Bangla Sign Language Deep Learning Mediapipe Hand Gesture Recognition Assistive Technology Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version 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-8949739","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":595837065,"identity":"8677b0e1-599c-4e8e-875d-8be287953b68","order_by":0,"name":"Din Mohammad Toufik","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABFElEQVRIiWNgGAWjYDCCw0DM2ABmGgCxHI8EO4hnYEGUFkMgZcwjwXMApEUCt5YDaFoYJCQSQBzcWviO8x7d8HGHXbRu++HtDz62GchIznx+dcOPAgkG/vbuBGxaJA/zpd2ceSY5d9uZtMLGmW0GPNLSOWU3e4AOkzhzdgM2LQaHecxu87Yx5247kGPYzNv2h0dOOiftBg9Qi4FELj4t9bnbzr8BaTHgkZM8k3bzD2Eth3O33ciBaJGWYD92G58tkkAtN2e2HQdqeVY4c8Y5Ax7Jnhy22zIGEjy4/MJ3/ozZjY9t1UCHJW/48KHMwF7i+PFnN9/8sZHjb+/FqgUb4DEAk8QqBwH2B6SoHgWjYBSMguEPAHmoaZAtk9TJAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0009-0005-5603-874X","institution":"Metopolitan University, Sylhet, Bangladesh","correspondingAuthor":true,"prefix":"","firstName":"Din","middleName":"Mohammad","lastName":"Toufik","suffix":""},{"id":595837066,"identity":"1c821654-925a-477f-a435-6781e69be8a7","order_by":1,"name":"Sumayae Binata Khaiat","email":"","orcid":"","institution":"Sylhet Agricultural University, Bangladesh","correspondingAuthor":false,"prefix":"","firstName":"Sumayae","middleName":"Binata","lastName":"Khaiat","suffix":""}],"badges":[],"createdAt":"2026-02-23 18:00:25","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8949739/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8949739/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103421116,"identity":"5f62acd6-9bd1-4a69-9525-1a2808b0861d","added_by":"auto","created_at":"2026-02-25 13:16:52","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":714516,"visible":true,"origin":"","legend":"","description":"","filename":"BdSLSystemManuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8949739/v1_covered_7d9f28d8-1cdf-4173-8cc9-628ef5732cca.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eA Reliable Bangla Sign Language Recognition System Using MediaPipe and LSTM Networks\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Metropolitan University,Sylhet, Bangladesh","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"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":"Bangla Sign Language, Deep Learning, Mediapipe, Hand Gesture Recognition, Assistive Technology","lastPublishedDoi":"10.21203/rs.3.rs-8949739/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8949739/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGenerally, the traditional mode of communication among Bangladesh's speech and hearing challenged patients is through the mode referred to as \"Bangla Sign Language\" (abbreviated as \"BdSL\"). Though the mode exists as an effective bridge to connect and communicate among patients and the general public, the lack of an appropriate interpretation mechanism has led to an enormous communication problem among patients and the public. In the present study, an efficient and \"real-time\" mode referred to as \"Bangla Sign Language Recognition\" (acronym \"BSLR\"), where \"MediaPipe\" and \"Long Short-Term Memory\" (acronym \"LSTM\" networks) are utilized, is proposed as follows: Instead of living up to the usual and more cumbersome \"Convolutional Neural Network\" (acronym \"CNN\" networks) based methods requiring image processing, skeleton \"keypoints\" have been considered for efficient \"comprehension\" and \"perception\".\u003c/p\u003e\n\u003cp\u003eIn order to ensure that the systems have adequate resistance from backgrounds and lighting changes, our team used holistics from Mediapipe to obtain relevant features (x, y, z coordinates) from both hands and body postures. To classify the gesture set, relevant features extracted from both hands and postures have been used as input to the specially developed LSTM scheme to classify the gesture set input to the specially developed LSTM scheme. In order to test and confirm our system, an entire dataset was prepared consisting of 50 BdSL signals (including alphabets and phrases) from 15 participants in 7,500 video clips.\u003c/p\u003e\n\u003cp\u003eThe proposed design has a training accuracy of 99.2% and a testing accuracy of 98.5%, as supported by the experimental results. Further, it has an inference latency as low as 40 ms, which allows it to be useful for the deployment of real-time communication systems.\u003c/p\u003e\n\u003cp\u003eThis work validates the assertion that for continuous Bangla Sign Language recognition, the combination of the effective features extracted by MediaPipe and the powerful sequence learning capabilities offered by LSTMs far surpass the state-of-the-art models based on static images.\u003c/p\u003e","manuscriptTitle":"A Reliable Bangla Sign Language Recognition System Using MediaPipe and LSTM Networks","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-25 13:16:33","doi":"10.21203/rs.3.rs-8949739/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":"00d286eb-7139-4b9a-acec-7a93e8dd1b69","owner":[],"postedDate":"February 25th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":63520940,"name":"Cell Communication and Signaling"}],"tags":[],"updatedAt":"2026-02-25T13:16:34+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-25 13:16:33","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8949739","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8949739","identity":"rs-8949739","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}