Deep Learning Approaches for Continuous Sign Language Recognition: A Comprehensive Review
Sign language uses hand gestures as a visual mode of communication, along with body actions and facial expressions. Due to the increasing incidence of hearing deficiencies, the field of Continuous Sign Language Recognition (CSLR) has seen a considerable increase in research, which involves identifyi...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10937713/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849737007977201664 |
|---|---|
| author | Asma Khan Seyong Jin Geon-Hee Lee Gul E. Arzu L. Minh Dang Tan N. Nguyen Woong Choi Hyeonjoon Moon |
| author_facet | Asma Khan Seyong Jin Geon-Hee Lee Gul E. Arzu L. Minh Dang Tan N. Nguyen Woong Choi Hyeonjoon Moon |
| author_sort | Asma Khan |
| collection | DOAJ |
| description | Sign language uses hand gestures as a visual mode of communication, along with body actions and facial expressions. Due to the increasing incidence of hearing deficiencies, the field of Continuous Sign Language Recognition (CSLR) has seen a considerable increase in research, which involves identifying consecutive signs in video streams without previous information of their sequential limitations. However, existing research often lacks a unified framework for integrating spatial, temporal, and alignment approaches, while critical challenges such as real-time processing, diverse datasets, and signer variability remain unresolved. This survey uniquely contributes by presenting a novel framework that integrates these dimensions into a unified taxonomy for CSLR systems. It critically analyzes numerous studies, organizing them into a comprehensive taxonomy covering aspects such as sign language, data collection, input method, gesture signals, identification methods, applied data collections, and comprehensive efficiency. The article further categorizes deep-learning CSLR models according to spatial, temporal, and alignment approaches, highlighting their benefits and drawbacks. Furthermore, it explores various research aspects, such as the challenges of CSLR, the significance of nonverbal elements in CSLR systems, and the gaps in the body of current research. By emphasizing the role of real-time processing and diverse datasets, this survey provides actionable insights for advancing CSLR systems in practical applications. This classification serves as a helpful tool for researchers developing and organizing cutting-edge CSLR methods. The study highlights the effectiveness of deep learning systems in capturing different sign language signals. On the other hand, several challenges remain, such as the need for diverse, naturalistic datasets, improved signer diversity, and real-time CSLR systems. Addressing these gaps will be essential for advancing CSLR’s real-world applications and developing more robust, efficient models for the future. The conclusions give a wider Comprehension of sign language recognition and set the groundwork for future studies focused on addressing the current challenges and issues in this developing area. |
| format | Article |
| id | doaj-art-1f3d747a31bc4e96ab43f999c75bf48c |
| institution | DOAJ |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-1f3d747a31bc4e96ab43f999c75bf48c2025-08-20T03:07:05ZengIEEEIEEE Access2169-35362025-01-0113555245554410.1109/ACCESS.2025.355404610937713Deep Learning Approaches for Continuous Sign Language Recognition: A Comprehensive ReviewAsma Khan0https://orcid.org/0009-0000-5604-9813Seyong Jin1https://orcid.org/0009-0004-4537-8581Geon-Hee Lee2Gul E. Arzu3L. Minh Dang4Tan N. Nguyen5Woong Choi6https://orcid.org/0000-0001-6635-8951Hyeonjoon Moon7https://orcid.org/0000-0001-7668-3838Department of Computer Science and Engineering, Sejong University, Seoul, Republic of KoreaDepartment of Artificial Intelligence, Sejong University, Seoul, Republic of KoreaDepartment of Artificial Intelligence, Sejong University, Seoul, Republic of KoreaDepartment of Computer Science and Engineering, Sejong University, Seoul, Republic of KoreaInstitute of Research and Development, Duy Tan University, Da Nang, VietnamDepartment of Architectural Engineering, Sejong University, Gwangjin, Seoul, Republic of KoreaDivision of ICT Convergence Engineering, Kangnam University, Yongin, Republic of KoreaDepartment of Computer Science and Engineering, Sejong University, Seoul, Republic of KoreaSign language uses hand gestures as a visual mode of communication, along with body actions and facial expressions. Due to the increasing incidence of hearing deficiencies, the field of Continuous Sign Language Recognition (CSLR) has seen a considerable increase in research, which involves identifying consecutive signs in video streams without previous information of their sequential limitations. However, existing research often lacks a unified framework for integrating spatial, temporal, and alignment approaches, while critical challenges such as real-time processing, diverse datasets, and signer variability remain unresolved. This survey uniquely contributes by presenting a novel framework that integrates these dimensions into a unified taxonomy for CSLR systems. It critically analyzes numerous studies, organizing them into a comprehensive taxonomy covering aspects such as sign language, data collection, input method, gesture signals, identification methods, applied data collections, and comprehensive efficiency. The article further categorizes deep-learning CSLR models according to spatial, temporal, and alignment approaches, highlighting their benefits and drawbacks. Furthermore, it explores various research aspects, such as the challenges of CSLR, the significance of nonverbal elements in CSLR systems, and the gaps in the body of current research. By emphasizing the role of real-time processing and diverse datasets, this survey provides actionable insights for advancing CSLR systems in practical applications. This classification serves as a helpful tool for researchers developing and organizing cutting-edge CSLR methods. The study highlights the effectiveness of deep learning systems in capturing different sign language signals. On the other hand, several challenges remain, such as the need for diverse, naturalistic datasets, improved signer diversity, and real-time CSLR systems. Addressing these gaps will be essential for advancing CSLR’s real-world applications and developing more robust, efficient models for the future. The conclusions give a wider Comprehension of sign language recognition and set the groundwork for future studies focused on addressing the current challenges and issues in this developing area.https://ieeexplore.ieee.org/document/10937713/Continuous sign language recognitiondeep learninghand gesture recognition (HGR)computer vision |
| spellingShingle | Asma Khan Seyong Jin Geon-Hee Lee Gul E. Arzu L. Minh Dang Tan N. Nguyen Woong Choi Hyeonjoon Moon Deep Learning Approaches for Continuous Sign Language Recognition: A Comprehensive Review IEEE Access Continuous sign language recognition deep learning hand gesture recognition (HGR) computer vision |
| title | Deep Learning Approaches for Continuous Sign Language Recognition: A Comprehensive Review |
| title_full | Deep Learning Approaches for Continuous Sign Language Recognition: A Comprehensive Review |
| title_fullStr | Deep Learning Approaches for Continuous Sign Language Recognition: A Comprehensive Review |
| title_full_unstemmed | Deep Learning Approaches for Continuous Sign Language Recognition: A Comprehensive Review |
| title_short | Deep Learning Approaches for Continuous Sign Language Recognition: A Comprehensive Review |
| title_sort | deep learning approaches for continuous sign language recognition a comprehensive review |
| topic | Continuous sign language recognition deep learning hand gesture recognition (HGR) computer vision |
| url | https://ieeexplore.ieee.org/document/10937713/ |
| work_keys_str_mv | AT asmakhan deeplearningapproachesforcontinuoussignlanguagerecognitionacomprehensivereview AT seyongjin deeplearningapproachesforcontinuoussignlanguagerecognitionacomprehensivereview AT geonheelee deeplearningapproachesforcontinuoussignlanguagerecognitionacomprehensivereview AT gulearzu deeplearningapproachesforcontinuoussignlanguagerecognitionacomprehensivereview AT lminhdang deeplearningapproachesforcontinuoussignlanguagerecognitionacomprehensivereview AT tannnguyen deeplearningapproachesforcontinuoussignlanguagerecognitionacomprehensivereview AT woongchoi deeplearningapproachesforcontinuoussignlanguagerecognitionacomprehensivereview AT hyeonjoonmoon deeplearningapproachesforcontinuoussignlanguagerecognitionacomprehensivereview |