RPC-Net Dataset. Simultaneous HD-sEMG Recordings on the Forearm and angles of a 29-DOF Hand Kinematic Model

The dataset in this repository comprises data acquired during the doctoral research project of Giovanni Rolandino at the Nuffield Department of Surgical Sciences, University of Oxford. Five sub-datasets make up the repository: DS1: Simultaneous acquisition of high-density surface electromyography (HD-sEMG) signals from the forearm and hand position kinematics. Data were recorded from 12 healthy subjects while they cycled through 16 hand poses. HD-sEMG was acquired with traditional gel electrode arrays. DS2: A similar protocol to DS1 was followed, but the HD-sEMG was acquired using a novel dry-electrode array. This dataset includes 16 subjects. Whereas DS1 included data from a single session for each subject, DS2 includes two sessions, acquired hours to days apart; these sessions are identified as s1 and s2. DS3: This subset consists of two parts. DS3.a repeats the protocol used in DS2 with 4 subjects, introducing repositioning between trials. DS3.b includes the results of the real-time assessment of RPC-Net, a shallow neural network trained with data from DS3.a to estimate hand position from HD-sEMG activity. DS3.b contains the real-time output recorded during prompt-matching tasks and the corresponding targets. DS4: This subset includes data related to the assessment of RFC-Net, a shallow neural network designed to estimate hand position from neck muscle activation. Experiments were performed on 8 healthy participants and 8 participants with tetraplegia. DS4.a includes the data used for training the network, while DS4.b includes data from the testing phase of the algorithm. DS4.b.s1 includes results from a cursor control task, and DS4.b.s2 includes results from a virtual hand control task. AD1: Additional data related to the electrical validation of the dry-electrode array. Code for processing the data in this repository is available on Dropbox:https://www.dropbox.com/scl/fo/nkvbse7evo0k8ou1utn7i/AMDh_MOZQJ6gCwDXGPadmZ0?rlkey=ynoix3anpc81v24hogn3fymb4&st=xrqx07q2&dl=0 For additional information, readers are referred to the original papers detailing acquisition protocols and processing procedures: 1) G. Rolandino, M. Gagliardi, T. Martins, G. L. Cerone, B. Andrews, J. J. FitzGerald. Developing RPC-Net: Leveraging High-Density Electromyography and Machine Learning for Improved Hand Position Estimation. IEEE Transactions on Biomedical Engineering, 71(5):1617-1627, May 2024. doi:10.1109/TBME.2023.3346192. 2) G. Rolandino, C. Zangrandi, T. Vieira, G. L. Cerone, B. Andrews, J. J. FitzGerald. HDE-Array: Development and Validation of a New Dry Electrode Array Design to Acquire HD-sEMG for Hand Position Estimation. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 32:4004-4013, 2024. doi:10.1109/TNSRE.2024.3490796. 3) G. Rolandino, T. Vieira, G. L. Cerone, B. Andrews, J. J. FitzGerald. Performance of a ML-Based 3-DoF Kinematic Model in Estimating Hand Position from High-Density EMG. Presented at IFESS Conference, Bath, UK, September 2024. 4) G. Rolandino, L. Lion, T. Vieira, G. L. Cerone, B. Andrews, J. J. FitzGerald. Artificial Neural Networks for HD-sEMG-Based Hand Position Estimation: Addressing Inter- and Intra-Subject Variability. Submitted to IEEE Transactions on Neural Systems and Rehabilitation Engineering, July 2025. 5) G. Rolandino, G. Parisi, T. Vieira, G. L. Cerone, B. Andrews, J. J. FitzGerald. Real-Time Hand Kinematic Estimation with HD-sEMG and Artificial Neural Networks: Feasibility and Effects of Multi-Subject Training and Visual Feedback. Submitted to IEEE Transactions on Neural Systems and Rehabilitation Engineering, July 2025. 6) G. Rolandino, V. Taboni Lisboa, T. Vieira, A. Cliquet Jr., B. Andrews, J. J. FitzGerald. HD-sEMG-Based Control Using Neck Muscles and Shallow Neural Networks: Assessing Performance in Rehabilitation-Oriented Tasks. Submitted to IEEE Transactions on Neural Systems and Rehabilitation Engineering, July 2025. This dataset benefited from the support of all listed authors and arose from collaborations between the Oxford Neural Interfacing Group; LISiN (Politecnico di Torino, Turin, Italy); the Department of Orthopedics, Rheumatology and Traumatology (University of Campinas, SP, Brazil); and the Oxford Robotics Institute (University of Oxford, Oxford, UK). Part of this work was funded by the John Fell Oxford University Press Research Fund. The corresponding author is available for questions or clarification at [e-mail].