Low-Latency Neural Inference on an Edge Device for Real-Time Handwriting Recognition from EEG Signals

Low-Latency Neural Inference on an Edge Device for Real-Time Handwriting Recognition from EEG Signals | 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 Article Low-Latency Neural Inference on an Edge Device for Real-Time Handwriting Recognition from EEG Signals Ovishake Sen, Raghav Soni, Darpan Virmani, Akshar Parekh, Patrick Lehman, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7330202/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 20 Nov, 2025 Read the published version in Scientific Reports → Version 1 posted 12 You are reading this latest preprint version Abstract Brain–computer interfaces (BCIs) hold significant promise for restoring communication in individuals with severe motor or speech impairments. Imagined handwriting, as a form of motor imagery, offers an intuitive paradigm for character-level neural decoding. While invasive techniques such as electrocorticography (ECoG) offer high decoding accuracy, their surgical requirements pose clinical risks and hinder scalability. Non-invasive alternatives like electroencephalography (EEG) are safer and more accessible but suffer from low signal-to-noise ratio (SNR) and spatial resolution, limiting their effectiveness in high-resolution decoding. Here, we investigate how advanced machine learning, combined with informative feature extraction, can overcome these limitations—enabling EEG-based decoding performance that approaches invasive methods, while supporting real-time inference on edge devices. We present the first real-time, low-latency, high-accuracy system for decoding imagined handwriting from non-invasive EEG signals on a portable edge device. EEG data were collected from seven participants using a 32-channel headcap and preprocessed with bandpass filtering and artifact subspace reconstruction. We extracted 20 time-and frequency-domain features, then applied Pearson correlation coefficient-based feature selection to reduce latency while preserving accuracy. A hybrid architecture combining a Temporal Convolutional Network (TCN) and a multilayer perceptron(MLP) was trained on the extracted features and deployed on the NVIDIA Jetson TX2. The system achieved 83.64%±0.50%accuracy with 766.68 ms per-character inference latency. By selecting only four key features, the model incurred a minimal accuracy loss of less than 1%, while achieving a 4.93× reduction in inference latency (155.68 ms) compared to the full 20-feature set. These findings show that non-invasive EEG, combined with efficient feature and model design, can enable accurate, real-time neural decoding on low-power edge devices—paving the way for practical, portable BCIs. Biological sciences/Computational biology and bioinformatics Physical sciences/Engineering Physical sciences/Mathematics and computing Biological sciences/Neuroscience EEG decoding imagined handwriting real-time inference edge device BCI lightweight machine learning models Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 20 Nov, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 24 Aug, 2025 Reviews received at journal 24 Aug, 2025 Reviews received at journal 20 Aug, 2025 Reviewers agreed at journal 20 Aug, 2025 Reviewers agreed at journal 19 Aug, 2025 Reviewers agreed at journal 19 Aug, 2025 Reviewers agreed at journal 19 Aug, 2025 Reviewers invited by journal 19 Aug, 2025 Editor assigned by journal 19 Aug, 2025 Editor invited by journal 19 Aug, 2025 Submission checks completed at journal 14 Aug, 2025 First submitted to journal 14 Aug, 2025 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-7330202","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":504952311,"identity":"0fb20ea8-def7-42a8-a57b-fa1645ed541b","order_by":0,"name":"Ovishake Sen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYFACxgZmBoYECPsDg5wMkDIgXgvjDAZjHiK0MDDAtTDzEKOFXyK5+XNBTZocA//hY49t2wx4GNibt0ng0yI5I7FNesaxHGMGibR041yQFp5jZXi1GJw52MbMw1aR2CDBYyad2/aHh0Eix4yQlubPPP8q6hv4z3+TtgTZIv+GgJbjjQ3SvG05CQwMOWzSjCAtQOvw+6W9sU2aty/NsE0izUyy55wBDxtPWrEFPi38zOyPP/N8S5bn5z/8TOJHmYEcP/vhjTfwaYEDNgzGKBgFo2AUjALyAQA2BTs/jfRx/AAAAABJRU5ErkJggg==","orcid":"","institution":"University of Florida","correspondingAuthor":true,"prefix":"","firstName":"Ovishake","middleName":"","lastName":"Sen","suffix":""},{"id":504952312,"identity":"2cbd520b-fe0c-482a-8ad4-325507c21f91","order_by":1,"name":"Raghav Soni","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Raghav","middleName":"","lastName":"Soni","suffix":""},{"id":504952313,"identity":"02e67a83-3fd0-486e-bd2c-bdff9c8137ef","order_by":2,"name":"Darpan Virmani","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Darpan","middleName":"","lastName":"Virmani","suffix":""},{"id":504952315,"identity":"604b3ff4-d922-4a34-b3c7-8f83c1b59dd4","order_by":3,"name":"Akshar Parekh","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Akshar","middleName":"","lastName":"Parekh","suffix":""},{"id":504952317,"identity":"a327b6ed-4540-4cae-aa7d-b539f069ebd3","order_by":4,"name":"Patrick Lehman","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Patrick","middleName":"","lastName":"Lehman","suffix":""},{"id":504952318,"identity":"a4a8c0cb-e0da-43f5-9303-d4f7098f78eb","order_by":5,"name":"Sarthak Jena","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Sarthak","middleName":"","lastName":"Jena","suffix":""},{"id":504952323,"identity":"55835c4b-f262-403e-98f1-1c5a462748cb","order_by":6,"name":"Baibhab Chatterjee","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Baibhab","middleName":"","lastName":"Chatterjee","suffix":""}],"badges":[],"createdAt":"2025-08-08 21:23:07","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7330202/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7330202/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-24972-y","type":"published","date":"2025-11-20T15:57:43+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":96650468,"identity":"0cb863a5-d06b-41aa-97ec-87d7d0d0ff49","added_by":"auto","created_at":"2025-11-24 16:12:10","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":614773,"visible":true,"origin":"","legend":"","description":"","filename":"LowLatencyNeuralInferenceonanEdgeDeviceforRealTimeHandwritingRecognitionfromEEGSignals.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7330202/v1_covered_131f6fbf-19f9-4fad-a1f6-42e6783ae55a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Low-Latency Neural Inference on an Edge Device for Real-Time Handwriting Recognition from EEG Signals","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"EEG decoding, imagined handwriting, real-time inference, edge device, BCI, lightweight machine learning models","lastPublishedDoi":"10.21203/rs.3.rs-7330202/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7330202/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Brain–computer interfaces (BCIs) hold significant promise for restoring communication in individuals with severe motor or speech impairments. Imagined handwriting, as a form of motor imagery, offers an intuitive paradigm for character-level neural decoding. While invasive techniques such as electrocorticography (ECoG) offer high decoding accuracy, their surgical requirements pose clinical risks and hinder scalability. Non-invasive alternatives like electroencephalography (EEG) are safer and more accessible but suffer from low signal-to-noise ratio (SNR) and spatial resolution, limiting their effectiveness in high-resolution decoding. Here, we investigate how advanced machine learning, combined with informative feature extraction, can overcome these limitations—enabling EEG-based decoding performance that approaches invasive methods, while supporting real-time inference on edge devices. We present the first real-time, low-latency, high-accuracy system for decoding imagined handwriting from non-invasive EEG signals on a portable edge device. EEG data were collected from seven participants using a 32-channel headcap and preprocessed with bandpass filtering and artifact subspace reconstruction. We extracted 20 time-and frequency-domain features, then applied Pearson correlation coefficient-based feature selection to reduce latency while preserving accuracy. A hybrid architecture combining a Temporal Convolutional Network (TCN) and a multilayer perceptron(MLP) was trained on the extracted features and deployed on the NVIDIA Jetson TX2. The system achieved 83.64%±0.50%accuracy with 766.68 ms per-character inference latency. By selecting only four key features, the model incurred a minimal accuracy loss of less than 1%, while achieving a 4.93× reduction in inference latency (155.68 ms) compared to the full 20-feature set. These findings show that non-invasive EEG, combined with efficient feature and model design, can enable accurate, real-time neural decoding on low-power edge devices—paving the way for practical, portable BCIs.","manuscriptTitle":"Low-Latency Neural Inference on an Edge Device for Real-Time Handwriting Recognition from EEG Signals","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-27 06:24:04","doi":"10.21203/rs.3.rs-7330202/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-08-25T01:12:42+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-24T22:56:45+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-20T10:03:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"306564855198285914335666203038854540181","date":"2025-08-20T10:02:21+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"205330677657320978103987267194948247703","date":"2025-08-19T22:50:16+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"50599638322979517533464816952683603817","date":"2025-08-19T12:24:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"17956184430400379072027366950243630572","date":"2025-08-19T08:13:35+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-19T07:59:09+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-19T07:42:48+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-08-19T07:28:29+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-14T17:45:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-08-14T17:44:13+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"7eea04fa-9c6f-4bfc-af48-89e0362fe3d1","owner":[],"postedDate":"August 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":53633194,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":53633195,"name":"Physical sciences/Engineering"},{"id":53633196,"name":"Physical sciences/Mathematics and computing"},{"id":53633197,"name":"Biological sciences/Neuroscience"}],"tags":[],"updatedAt":"2025-11-24T16:08:02+00:00","versionOfRecord":{"articleIdentity":"rs-7330202","link":"https://doi.org/10.1038/s41598-025-24972-y","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-11-20 15:57:43","publishedOnDateReadable":"November 20th, 2025"},"versionCreatedAt":"2025-08-27 06:24:04","video":"","vorDoi":"10.1038/s41598-025-24972-y","vorDoiUrl":"https://doi.org/10.1038/s41598-025-24972-y","workflowStages":[]},"version":"v1","identity":"rs-7330202","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7330202","identity":"rs-7330202","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}