Physiological Signal Synthesis and Classification: Assessing Entropy Change and Other Physiological Signals
Description
This dataset accompanies the manuscript "Physiological Signal Synthesis and Classification: Assessing Entropy Change and Other Health Metrics". It contains a single-subject physiological dataset of 1,806 samples collected across seven tasks, featuring heart rate (HR), blood pressure (BP), temperature (T), and derived entropy values (ΔS = ΔHR × ΔBP / ΔT). In addition to the original data, the dataset includes 13 synthetic data variants generated using bootstrap resampling, variational autoencoders (VAEs), and generative adversarial networks (GANs), including WGAN-GP, CycleGAN, InfoGAN, and others. Each synthetic dataset is aligned in format and labeling with the original dataset to support reproducibility and direct classifier training. This dataset is suitable for research in synthetic data generation, biomedical machine learning, signal classification, and entropy modeling.
Files
Steps to reproduce
This dataset comprises synthetic physiological data, along with the code used to generate it and analyze its properties. Files Included: physiological_data.csv: Raw physiological data. synthetic_data_generation_paper_code.ipynb: Research code for data generation and analysis. syntheticData.zip: 63 datasets of synthetic data generated to bypass time-consuming data generation processes. Data Processing and Analysis: DataPipelineV2: Helper function to preprocess and add features to the raw data. Time-Saving Tip: Plots such as CDF, ROC, and PCA are time-consuming to generate. The outputs are available within the .ipynb file. Usage Notes: The code is prepared in a Colab notebook. If using Jupyter Notebook or converting it to a .py file, please change the directory of the data files and the helper function to reproduce the results.
Institutions
- Universiti Tunku Abdul Rahman