Signal Dataset and Machine Learning Codes for Exercise-Associated Motion Classification Using PVDF/PVA Piezoelectric Hydrogel Sensors

This dataset contains the signal data and machine learning codes generated for exercise-associated motion classification using wearable PVDF/PVA piezoelectric hydrogel sensors. The dataset supports the research reported in the manuscript entitled “All-polymer sea-island structured hydrogels with enhanced mechanical and piezoelectric properties for wearable fitness monitoring”. The signal dataset consists of piezoelectric output signals collected from wearable PVDF/PVA hydrogel sensors during exercise-related motion monitoring. A total of 1,200 bench-press movement trials were collected from 12 participants, including 600 standard movements and 600 non-standard movements. The dataset was established for subject-independent motion classification using a 12-fold leave-one-subject-out (LOSO) cross-validation strategy. This repository also provides the corresponding machine learning codes, including the Long Short-Term Memory (LSTM) neural network model and conventional machine learning algorithms (RBF-SVM and Random Forest) used for performance comparison. The provided codes include data preprocessing, model training, testing procedures, and evaluation workflows. The dataset and source codes enable reproducibility of the machine learning analysis and provide resources for further research on wearable piezoelectric sensors, biomechanical signal processing, and intelligent motion monitoring.