Add WESAD Dataset for Wearable Stress Detection

[ChengcenZhou]

Contributor Info

• Name: Chenngcen Zhou
• Email: [email protected]

Type of Contribution

• New dataset (WESAD: Wearable Stress and Affect Detection)
• New tasks (stress_detection_wesad_fn, binary and multiclass variants)
• New unit tests

What's WESAD Dataset?

This PR integrates the WESAD dataset [Schmidt et al., 2018] into PyHealth for wearable stress detection research. It adds:

• A WESADDataset class for loading Empatica E4 wrist-worn sensor data (EDA, BVP, ACC, TEMP) from 15 subjects
• Task functions for binary stress detection (stress vs non-stress) and 3-class classification (baseline, stress, amusement)
• Unit tests with synthetic data generation

The implementation follows the contributing guidelines (PEP8, Google-style docstrings, and documented function signatures).

Files to Review

Modified:

• pyhealth/datasets/__init__.py: register WESADDataset
• pyhealth/tasks/__init__.py: register stress_detection_wesad_fn

New:

• pyhealth/datasets/wesad.py: WESADDataset class implementation
• pyhealth/tasks/stress_detection_wesad.py: Task functions for stress detection
• tests/test_wesad.py: Unit tests for dataset and tasks

Dataset Info

Item	Details
Subjects	15 (S2-S17, excluding S1, S12)
Device	Empatica E4 wristband
Signals	EDA (4Hz), BVP (64Hz), ACC (32Hz), TEMP (4Hz)
Labels	Baseline, Stress, Amusement
Source	UCI ML Repository / Kaggle

Reference

• Paper Title: Schmidt, P., Reiss, A., Duerichen, R., Marberger, C., & Van Laerhoven, K. "Introducing WESAD, a Multimodal Dataset for Wearable Stress and Affect Detection", ICMI 2018
• Paper Link: https://doi.org/10.1145/3242969.3242985
• Dataset Link: https://archive.ics.uci.edu/ml/datasets/WESAD

How to Use

from pyhealth.datasets import WESADDataset
from pyhealth.tasks import stress_detection_wesad_fn

# Load dataset
dataset = WESADDataset(root="/path/to/WESAD/")

# Apply task
dataset = dataset.set_task(stress_detection_wesad_fn)

# Access samples
print(f"Total samples: {len(dataset.samples)}")