A Novel Hybrid GRU-LSTM Model for Robust Speech Recognition of Indian Languages in Noisy Environments

A Novel Hybrid GRU-LSTM Model for Robust Speech Recognition of Indian Languages in Noisy Environments | 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 Novel Hybrid GRU-LSTM Model for Robust Speech Recognition of Indian Languages in Noisy Environments Astha Gupta, Rakesh Kumar, Meenu Gupta, Durgesh Srivastava, Amit Garg This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9454210/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract Speech is a naturally occurring phenomenon that depends on the speaker, surroundings, language, accent, and tone. During the speech, the waveforms change over time, making it challenging for Automatic Speech Recognition (ASR) system to analyze and identify these signals in a noisy environment. In this work, A hybrid model GRU-LSTM a combination of Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM), is proposed to recognize the speeches of the five Indian languages (i.e., Bangla, Malayalam, Odia, Telugu, and Urdu). A total of 8682 speech samples of the abovementioned languages are used from which various features are extracted using techniques such as Spectrogram, Short Term Fourier Transform (STFT), Spectral Rolloff, Spectral Bandwidth, Mel-Frequency Cepstral Coefficient (MFCC), and Chroma feature. In addition, the proposed model is tuned at three different learning rates such as 0.1, 0.001, and 0.0001, where each learning rate is analyzed over 0.2, 0.3, and 0.4 dropouts, respectively, to obtain the highest accuracy. By this, the highest validation accuracy of 97.02% with a loss of 0.41 is achieved by tuning the model at 0.001 learning rate at 0.3 dropout. Next, the performance of proposed and SOTA (State of The Art) models for each language is evaluated based on Training Accuracy, Loss, Validation Accuracy, and Loss with Precision, Recall, F1-Score, and Confusion Matrix. Finally, an analogy of the proposed with other SOTA models is also depicted. The performance of each model for each language is evaluated based on Precision, Recall, F1-Score, Accuracy, Loss, and Confusion Matrix. Finally, it is concluded that the proposed model attained the validation accuracy of 96.56%, 96.56%, 98.56%, 94.59% and 98.95% is achieved with 0.469, 0.599, 0.469,0.469 and 0.059 loss for Bangla, Malayalam, Odia, Telugu and Urdu languages. ASR Hybrid GRU-LSTM Feature Extraction Spectrogram STFT Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 07 May, 2026 Editor assigned by journal 22 Apr, 2026 Submission checks completed at journal 22 Apr, 2026 First submitted to journal 18 Apr, 2026 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-9454210","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":641579768,"identity":"89478253-7589-470b-8f49-51dba17a1b39","order_by":0,"name":"Astha Gupta","email":"","orcid":"","institution":"Chitkara University Institute of Engineering \u0026 Technology, Chitkara University","correspondingAuthor":false,"prefix":"","firstName":"Astha","middleName":"","lastName":"Gupta","suffix":""},{"id":641579773,"identity":"ef343c59-0bf4-4af9-b98c-5d41a87a3323","order_by":1,"name":"Rakesh Kumar","email":"","orcid":"","institution":"Chandigarh University","correspondingAuthor":false,"prefix":"","firstName":"Rakesh","middleName":"","lastName":"Kumar","suffix":""},{"id":641579775,"identity":"7b912551-a95a-46ab-b14a-de91020b0307","order_by":2,"name":"Meenu Gupta","email":"","orcid":"","institution":"Chandigarh University","correspondingAuthor":false,"prefix":"","firstName":"Meenu","middleName":"","lastName":"Gupta","suffix":""},{"id":641579780,"identity":"3762824c-c631-414d-9939-b6cad570eda1","order_by":3,"name":"Durgesh Srivastava","email":"","orcid":"","institution":"Chitkara University Institute of Engineering \u0026 Technology, Chitkara University","correspondingAuthor":false,"prefix":"","firstName":"Durgesh","middleName":"","lastName":"Srivastava","suffix":""},{"id":641579784,"identity":"f09c7a2a-e0e3-4694-b24d-8fdb2fb0b738","order_by":4,"name":"Amit Garg","email":"data:image/png;base64,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","orcid":"","institution":"Manipal University Jaipur","correspondingAuthor":true,"prefix":"","firstName":"Amit","middleName":"","lastName":"Garg","suffix":""}],"badges":[],"createdAt":"2026-04-18 05:08:27","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9454210/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9454210/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109432839,"identity":"e8865847-8b82-4f00-b658-bf4e18483aa2","added_by":"auto","created_at":"2026-05-18 05:33:41","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1306582,"visible":true,"origin":"","legend":"","description":"","filename":"ManuscriptIndian.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9454210/v1_covered_51d55288-2bf8-4867-bac8-13f69dcddd88.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Novel Hybrid GRU-LSTM Model for Robust Speech Recognition of Indian Languages in Noisy Environments","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-computing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Computing](https://link.springer.com/journal/10791)","snPcode":"10791","submissionUrl":"https://submission.springernature.com/new-submission/10791/3","title":"Discover Computing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"ASR, Hybrid GRU-LSTM, Feature Extraction, Spectrogram, STFT","lastPublishedDoi":"10.21203/rs.3.rs-9454210/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9454210/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eSpeech is a naturally occurring phenomenon that depends on the speaker, surroundings, language, accent, and tone. During the speech, the waveforms change over time, making it challenging for Automatic Speech Recognition (ASR) system to analyze and identify these signals in a noisy environment. In this work, A hybrid model GRU-LSTM a combination of Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM), is proposed to recognize the speeches of the five Indian languages (i.e., Bangla, Malayalam, Odia, Telugu, and Urdu). A total of 8682 speech samples of the abovementioned languages are used from which various features are extracted using techniques such as Spectrogram, Short Term Fourier Transform (STFT), Spectral Rolloff, Spectral Bandwidth, Mel-Frequency Cepstral Coefficient (MFCC), and Chroma feature. In addition, the proposed model is tuned at three different learning rates such as 0.1, 0.001, and 0.0001, where each learning rate is analyzed over 0.2, 0.3, and 0.4 dropouts, respectively, to obtain the highest accuracy. By this, the highest validation accuracy of 97.02% with a loss of 0.41 is achieved by tuning the model at 0.001 learning rate at 0.3 dropout. Next, the performance of proposed and SOTA (State of The Art) models for each language is evaluated based on Training Accuracy, Loss, Validation Accuracy, and Loss with Precision, Recall, F1-Score, and Confusion Matrix. Finally, an analogy of the proposed with other SOTA models is also depicted. The performance of each model for each language is evaluated based on Precision, Recall, F1-Score, Accuracy, Loss, and Confusion Matrix. Finally, it is concluded that the proposed model attained the validation accuracy of 96.56%, 96.56%, 98.56%, 94.59% and 98.95% is achieved with 0.469, 0.599, 0.469,0.469 and 0.059 loss for Bangla, Malayalam, Odia, Telugu and Urdu languages.\u003c/p\u003e","manuscriptTitle":"A Novel Hybrid GRU-LSTM Model for Robust Speech Recognition of Indian Languages in Noisy Environments","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-18 05:33:36","doi":"10.21203/rs.3.rs-9454210/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2026-05-07T16:13:02+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-22T14:11:29+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-22T14:10:49+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Computing","date":"2026-04-18T04:53:33+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-computing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Computing](https://link.springer.com/journal/10791)","snPcode":"10791","submissionUrl":"https://submission.springernature.com/new-submission/10791/3","title":"Discover Computing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"feb1c175-dd1b-48a8-8129-74f0159200f8","owner":[],"postedDate":"May 18th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewersInvited","content":"8","date":"2026-05-07T16:13:02+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-18T05:33:37+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-18 05:33:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9454210","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9454210","identity":"rs-9454210","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}