Synthetic Health Sensor Data From Wearables With Stress Labels

Data stems from a adapted pipline of the models found in the paper:  Generating Synthetic Health Sensor Data for Privacy-Preserving Wearable Stress Detection
Please cite this paper as reference.

Both are synthetic time-series datasets created from training GANs on the WESAD dataset.
Each set features data for 10,000 subjects instead of just 15 found in the original.
An important difference to the original is the reduction of labels to only: stress and non-stress.
One is generated using a CGAN and the other a DGAN.

Used for testing re-identification attacks in: Slice it up: Unmasking User Identities in Smartwatch Health Data