Leveraging the Sequence of Activities to Enhance Sensor-Based Human Activity Recognition

Leveraging the Sequence of Activities to Enhance Sensor-Based Human Activity Recognition | 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 Leveraging the Sequence of Activities to Enhance Sensor-Based Human Activity Recognition AbdulRahman Baraka, Mohd Halim Noor This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9515041/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 Human activity recognition using wearable sensors has gained significant attention in recent research due to its use in numerous applications. In real-world scenarios, human activities are inherently sequential and context-dependent, where preceding actions often influence or constrain subsequent ones, particularly in transitional movements. For example, if the current activity is sitting, the next activity would be standing or lying down, not running or climbing stairs. Therefore, understanding the natural sequence of human activities helps accurately identify the current activity by prioritizing certain activities and excluding others. However, current deep learning models focus on identifying the current activity based on the spatial and temporal features of the windows, without explicitly learning the natural sequence of actions. In this paper, a deep learning architecture is designed to improve activity recognition by explicitly capturing the natural temporal sequence of human activities. By knowing the sequence of preceding activities, the model learns sequential dependencies, enabling more accurate recognition of the current activity using the knowledge of the previous activity. The proposed model leverages both local and temporal dependencies across adjacent windows through two pipelines. The first directly classifies the current activity from the current window, while the second utilizes the previous window to infer the current activity based on learning the activity sequence from the local and temporal context. A cross-attention mechanism is integrated within these pipelines by aligning current features with both past and present contexts, thereby refining the recognition process. A probability-based fusion strategy combines the outputs of both pipelines. Experimental evaluations on two publicly available datasets demonstrate that the proposed model significantly improves classification performance for both basic and transitional activities. Moreover, it outperforms state-of-the-art approaches, underscoring the critical role of context awareness and activity sequencing in advancing human activity recognition. The code is publicly available at: https://github.com/abaraka2020/SeqHAR Human Activity Recognition Deep Learning Cross-Attention Sequence Activity Transitional Activity Full Text Additional Declarations No competing interests reported. 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-9515041","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":633885696,"identity":"86762772-e2c8-422b-8808-34429b62e7b9","order_by":0,"name":"AbdulRahman Baraka","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAv0lEQVRIiWNgGAWjYDACCRBxgCGBH8JlJkGLZAPJWgwOEKuFf3bzsw8/ztjkGR9PPibBUGGd2EDQkjvHjGf23EgrNjvzLE2C4Uw6YS0MNxKMGXg+HE7cdiPHTIKx7TBhLfI30j8z/vnwP3HzDJCWf0RoMbiRY8zMc+NA4gYJkJYGIrQY3jlTzCxzJjlxxplnyRYJx9KNCWqRu92+mfHNMbvE/vbkgzc+1FjLEtSCBBLAiCRAqvpRMApGwSgYMQAAQ71FfL/3gdcAAAAASUVORK5CYII=","orcid":"","institution":"Al-Quds Open University","correspondingAuthor":true,"prefix":"","firstName":"AbdulRahman","middleName":"","lastName":"Baraka","suffix":""},{"id":633885697,"identity":"5c2023e7-1fbd-43b1-bcc7-da34dbf780e2","order_by":1,"name":"Mohd Halim Noor","email":"","orcid":"","institution":"Universiti Sains Malaysia","correspondingAuthor":false,"prefix":"","firstName":"Mohd","middleName":"Halim","lastName":"Noor","suffix":""}],"badges":[],"createdAt":"2026-04-24 09:38:45","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9515041/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9515041/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108492378,"identity":"0f4299e9-6f65-4934-9ccb-2ab9fafd3dba","added_by":"auto","created_at":"2026-05-05 09:57:37","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":634697,"visible":true,"origin":"","legend":"","description":"","filename":"SeqHAR52026.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9515041/v1_covered_c99e1a18-dcf5-4799-a81c-80210b3bc5f4.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Leveraging the Sequence of Activities to Enhance Sensor-Based Human Activity Recognition","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"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":"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":"Human Activity Recognition, Deep Learning, Cross-Attention, Sequence Activity, Transitional Activity","lastPublishedDoi":"10.21203/rs.3.rs-9515041/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9515041/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eHuman activity recognition using wearable sensors has gained significant attention in recent research due to its use in numerous applications. In real-world scenarios, human activities are inherently sequential and context-dependent, where preceding actions often influence or constrain subsequent ones, particularly in transitional movements. For example, if the current activity is sitting, the next activity would be standing or lying down, not running or climbing stairs. Therefore, understanding the natural sequence of human activities helps accurately identify the current activity by prioritizing certain activities and excluding others. However, current deep learning models focus on identifying the current activity based on the spatial and temporal features of the windows, without explicitly learning the natural sequence of actions. In this paper, a deep learning architecture is designed to improve activity recognition by explicitly capturing the natural temporal sequence of human activities. By knowing the sequence of preceding activities, the model learns sequential dependencies, enabling more accurate recognition of the current activity using the knowledge of the previous activity. The proposed model leverages both local and temporal dependencies across adjacent windows through two pipelines. The first directly classifies the current activity from the current window, while the second utilizes the previous window to infer the current activity based on learning the activity sequence from the local and temporal context. A cross-attention mechanism is integrated within these pipelines by aligning current features with both past and present contexts, thereby refining the recognition process. A probability-based fusion strategy combines the outputs of both pipelines. Experimental evaluations on two publicly available datasets demonstrate that the proposed model significantly improves classification performance for both basic and transitional activities. Moreover, it outperforms state-of-the-art approaches, underscoring the critical role of context awareness and activity sequencing in advancing human activity recognition. The code is publicly available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/abaraka2020/SeqHAR\u003c/span\u003e\u003cspan address=\"https://github.com/abaraka2020/SeqHAR\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e","manuscriptTitle":"Leveraging the Sequence of Activities to Enhance Sensor-Based Human Activity Recognition","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-05 05:57:35","doi":"10.21203/rs.3.rs-9515041/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":"7d05ef3d-cdab-427e-b2a2-f0baa018245a","owner":[],"postedDate":"May 5th, 2026","published":true,"recentEditorialEvents":[{"type":"editorAssigned","content":"","date":"2026-05-04T06:29:02+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-05-04T06:28:17+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-05T05:57:35+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-05 05:57:35","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9515041","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9515041","identity":"rs-9515041","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}