Processed WESAD Dataset

This dataset contains preprocessed physiological signals from the WESAD (Wearable Stress and Affect Detection) dataset, processed for multimodal stress detection research. The dataset includes 86,583 signal windows from 15 subjects, with 16 hand-crafted physiological features extracted from electrocardiogram (ECG), electrodermal activity (EDA), and respiration signals. Each window represents 20 seconds of data sampled at 700 Hz, with binary labels indicating stress or non-stress states. Preprocessing included signal-specific filtering (high-pass filter at 0.8 Hz and 50 Hz notch filter for ECG; low-pass filter at 0.5 Hz for EDA; band-pass filter at 0.1-0.35 Hz for respiration), R-peak detection using the Pan-Tompkins algorithm, and feature extraction including heart rate variability metrics (RMSSD, SDNN, pNN50, LF/HF ratio), EDA statistics (mean, standard deviation, skin conductance responses), and respiration parameters (breathing rate, spectral power). The dataset is suitable for developing and benchmarking machine learning and deep learning models for stress detection, with data structured for Leave-One-Subject-Out cross-validation. This processed dataset is derived from the original WESAD dataset (Schmidt et al.).