Dataset: HD-sEMG and Shallow Neural Networks for Control of a Three Degree of Freedom Hand Kinematic Model in Healthy and Tetraplegic Subjects

This dataset contains high-density surface electromyography (HD-sEMG) signals and corresponding task performance metrics collected from a study investigating neck muscle activation for controlling virtual cursors and a 3-degree-of-freedom (DoF) hand kinematic model. EMG data were acquired from 10 tetraplegic and 8 healthy subjects using the HDE-Array in conjunction with the MEACS system. 

The dataset is structured into two parts: (1) training phase data (TrDS), where subjects performed predefined neck movements to train RTC-Net, a neural network for movement estimation, and (2) testing phase data, which includes two experimental tasks. In Part 1, subjects controlled three independent cursors in a task-based paradigm, with performance evaluated using Task Completion Score (TCS) and Normalized Distance (ND). In Part 2, subjects controlled a virtual hand kinematic model via neck muscle activation, with performance assessed using Mean Angular Distance (MAD) and Mean Distance (MD). Additionally, benchmark data (DS0) from a previous study involving forearm-based hand control are included for comparative analysis.

The dataset includes data acquired during the training phase and the results of the testing phase. All code (Matlab and Python) used for the paper is included. Readers are referred to the original paper (same title) for further details.