EEG, PPG, GSR Multimodal Physiological Signals for Rehabilitation Fatigue Detection (2025)

This dataset provides de-identified multimodal physiological signals and corresponding analysis codes dedicated to the research of rehabilitation fatigue detection, covering the whole research process from raw signal acquisition to machine learning model training. The dataset includes electroencephalogram (EEG) raw signals and combined photoplethysmography (PPG) and galvanic skin response (GSR) raw signals, all of which have been strictly de-identified in accordance with the ethical approval requirements (Ethics No.: HFUT20251208001H) to protect the privacy of research participants, with written informed consent obtained from all volunteers prior to data collection.The preprocessed and feature-extracted data files are also included, among which the EEG data is segmented into 3s, 5s, and 10s time scales for diverse research needs; in addition, the integrated preprocessed feature data of PPG-GSR and the fusion feature data of EEG-PPG-GSR are provided to support multi-modal signal analysis. Corresponding core MATLAB codes (developed and tested based on MATLAB R2018b) are attached, including training and validation codes for three classic machine learning models: K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Decision Tree (DT). All codes rely on MATLAB's Signal Processing Toolbox and Statistics and Machine Learning Toolbox, with independent running logic, which can directly load the dataset for experimental reproduction without additional plug-ins.This dataset belongs to the research field of systems physiology and biomedical engineering and integrates the technical application of information and computing sciences. It can provide reliable data and code support for the research on physiological signal processing, human fatigue assessment, and the application of machine learning in rehabilitation medicine, and effectively improve the reproducibility of relevant cross-disciplinary research. All data files are in .mat format, and the code files are in .m format, with a clear file structure and standard naming, which is convenient for academic research and experimental reuse.