[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-134750-en":3,"doc-seo-134750-105":30,"detail-sidebar-cat-0-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},134750,7971461740909,"Levi","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","A Closer Look at Skeleton-based Continuous Sign Language Recognition - ICCV Workshop paper","Skeleton-based Sign Language Recognition (SLR) offers robustness to visual noise, improved privacy, and lower computational overhead compared with video-based methods, yet building transferable continuous SLR systems from skeleton inputs remains challenging. This ICCV workshop paper performs a comprehensive study on representations, preprocessing, architectural choices, and cross-dataset transferability. The proposed approach achieves Top-1 performance on signer-independent recognition and unseen sentence generalization under the MSLR Track-1 challenge. Experiments validate skeleton-based modeling potential and provide practical guidance for designing efficient SLR systems.","This ICCV Workshop paper is the Open Access version, provided by the Computer Vision Foundation.  \nExcept for this watermark, it is identical to the accepted version; the final published version of the proceedings is available on IEEE Xplore.  \nA Closer Look at Skeleton-based Continuous Sign Language Recognition  \nYuecong Min 1,2 , Yifan Yang3 , Peiqi Jiao 1,2 , Zixi Nan2 , Xilin Chen 1,2 , 1 State Key Laboratory of AI Safety, Institute of Computing Technology,  \nChinese Academy of Sciences (CAS), Beijing, 100190, China  \n2University of Chinese Academy of Sciences, Beijing, 100049, China  \n3Huazhong University of Science and Technology, Wuhan, China  \nAbstract  \nSkeleton-based Sign Language Recognition (SLR) has emerged as a promising alternative to video-based approaches, offering robustness to visual noise, enhanced privacy, and reduced computational overhead, making it suitable for real-world deployment. However, effectively leveraging skeleton data and improving the generalization ability of skeleton-based continuous SLR methods remain open challenges. In this work, we conduct a comprehensive study of skeleton-based continuous SLR, focusing on input representations, preprocessing strategies, architectural designs, and cross-dataset transferability. The proposed method achieves Top-1 performance on both signerindependent recognition and unseen sentence generalization under the MSLR Track-1 challenge. Experimental results not only validate the potential of skeleton-based models but also provide practical insights for designing effective and efficient SLR systems. Our codes will be available at [https://github.com/VIPL-SLP/MSLR](https://github.com/VIPL-SLP/MSLR ICCV2025)[ ](https://github.com/VIPL-SLP/MSLR ICCV2025)[ICCV2025](https://github.com/VIPL-SLP/MSLR ICCV2025).  \n1. Introduction  \nSign Language Recognition (SLR) plays a vital role in enabling accessible communication for the Deaf community and facilitating human-computer interaction. While videobased approaches have made substantial progress recently, they remain vulnerable to challenges such as background clutter, lighting variability, and occlusion. In consequence, skeleton-based representations, extracted from human pose estimation, offer a compelling alternative by capturing essential kinematic cues from the hands, body, and face. These representations provide inherent robustness to visual noise, preserve user privacy, and significantly reduce computational and storage demands, making them well-suited for real-world, on-device deployment.  \nDriven by these advantages, skeleton-based SLR has attracted increasing attention. Early attempts often employ Graph Convolutional Networks (GCNs) to extract part-level  \nfeatures in isolated SLR [16], continuous SLR [17], and self-supervised learning [13] . Many recent sign language translation works [18, 21, 29] also utilize skeleton data as input for its efficiency, achieving performance comparable to video-based methods. In addition, the fusion of video and skeleton data has shown significant improvements in both recognition and translation [4, 31] . Despite this progress, the high degrees of freedom in hand and body movements pose significant challenges for effective feature extraction. Furthermore, the robustness of skeleton-based methods under more challenging scenarios, such as signer-independent recognition and unseen sentence generalization, remains underexplored. In this paper, we conduct a thoughtful exploration about the design choices of skeleton-based SLR, examining input representations, architectural choices, and cross-dataset transferability.  \nThe sparse and compact nature of skeleton data presents significant challenges in effectively capturing subtle spatiotemporal dynamics and discriminative motion patterns. To better exploit the usage of skeleton data in continuous SLR, we adopt CoSign [17] as our baseline model. CoSign adopts a group-wise graph convolutional module to capture independent signals and empl","cbCaimlIHlJuOcee","https://ap.wps.com/l/cbCaimlIHlJuOcee","pdf",1280827,1,7,"English","en",105,"# Introduction\n# Related Work\n## Continuous Sign Language Recognition","[{\"question\":\"What problem does skeleton-based continuous sign language recognition address?\",\"answer\":\"It targets continuous SLR by recognizing sequences from pose-derived skeleton data, aiming to gain robustness to visual noise and reduce computation compared with video-based pipelines.\"},{\"question\":\"Which aspects does the paper study to improve skeleton-based continuous SLR?\",\"answer\":\"It analyzes input representations, preprocessing strategies, architectural designs, and cross-dataset transferability, including framerate selection and pretraining datasets.\"},{\"question\":\"How does the proposed method perform on MSLR Track-1?\",\"answer\":\"The method achieves Top-1 performance on both signer-independent recognition and unseen sentence generalization in the MSLR Track-1 challenge.\"}]","A Closer Look at Skeleton-based Continuous Sign Language Recognition - ICCV Workshop paper | PDF",1787299020,18,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"a-closer-look-at-skeleton-based-continuous-sign-language-recognition-iccv-workshop-paper","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/a-closer-look-at-skeleton-based-continuous-sign-language-recognition-iccv-workshop-paper/134750/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-22","2026-08-21",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What problem does skeleton-based continuous sign language recognition address?","Question",{"text":76,"@type":77},"It targets continuous SLR by recognizing sequences from pose-derived skeleton data, aiming to gain robustness to visual noise and reduce computation compared with video-based pipelines.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"Which aspects does the paper study to improve skeleton-based continuous SLR?",{"text":81,"@type":77},"It analyzes input representations, preprocessing strategies, architectural designs, and cross-dataset transferability, including framerate selection and pretraining datasets.",{"name":83,"@type":74,"acceptedAnswer":84},"How does the proposed method perform on MSLR Track-1?",{"text":85,"@type":77},"The method achieves Top-1 performance on both signer-independent recognition and unseen sentence generalization in the MSLR Track-1 challenge.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":107,"slug":138},19,"General","general"]