Open Datasets for Upper Limb Motion-A Systematic Review

• Published in: IEEE access Vol. 13; pp. 74107 - 74127
• Main Authors: Santanche, Ranieri, Centrone, Antonia, Mattei, Lorenza, Di Puccio, Francesca
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: arm; Biomechanical databases; Biomechanics; Biosensors; Criteria; Data collection; Datasets; human movements; Inertial sensing devices; Limbs; Machine learning; Motion capture; Movement; open datasets; Optical sensors; Protocols; Sensor systems; Sensors; Sports; Systematic review; upper limb
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2025.3564350

Open biomechanical datasets play a critical role in advancing scientific research, particularly with the growing use of machine learning, which requires a large amount of data for training and validation. However, the availability and characteristics of such datasets, particularly in the context of upper limb motion, are still underexplored. This systematic review aims to identify, assess, and categorize existing open datasets related to upper limb motion, focusing on the tasks analyzed, subject demographics, and data collection methods employed. We conducted a comprehensive search across multiple databases, including Scopus, Zenodo, PubMed, Xplore, and Google Dataset Search, selecting publicly accessible datasets that focus on upper limb motion based on predefined inclusion and exclusion criteria. Additional datasets were identified through manual searches. A total of 63 datasets met the criteria for further analysis, and their quality was assessed based on subject types, data collection methods, and sample sizes. The majority of the datasets centered on activities of daily living, with 69.8% of the studies involving healthy subjects. Marker-based motion capture was the most common data collection method, though the use of markerless systems (e.g., inertial sensors, video cameras) is on the rise. Key limitations included the absence of standardized movement protocols and small subject sample sizes across studies. In contrast to lower limb gait analysis, there is no single dominant task for studying upper limb motion. Standardizing movement protocols and incorporating a broader range of daily activities into datasets could enhance the development of assistive technologies and rehabilitation programs. Future research should focus on creating more standardized, diverse datasets to improve the accuracy and generalizability of biomechanical analyses.