WESAD — Wearable Stress and Affect Detection [15 Subjects, 700 Hz]

Overview

WESAD (Wearable Stress and Affect Detection) is a publicly available multimodal dataset designed for stress and affect detection using wearable physiological sensors. Data were recorded from 15 subjects during a controlled lab study involving stress-inducing and neutral conditions, using both a chest-worn RespiBAN device and a wrist-worn Empatica E4.

The dataset captures a rich combination of physiological and motion signals including blood volume pulse (BVP), electrocardiogram (ECG), electrodermal activity (EDA), electromyogram (EMG), respiration, body temperature, and three-axis acceleration. Signals are sampled at frequencies ranging from 4 Hz to 700 Hz depending on the modality.

WESAD is widely cited in affective computing research and is considered a benchmark for wearable-based stress detection studies applying machine learning and deep learning approaches.

Column Schema

Key Statistics

• Total Subjects: 15
• Devices: RespiBAN (chest) + Empatica E4 (wrist)
• Sensor Modalities: 7 (ECG, EDA, EMG, RESP, TEMP, BVP, ACC)
• File Format: Pickle (Python) and CSV
• Sampling Rate: 4 Hz – 700 Hz depending on channel
• Time Period: 2018

Use Cases

• Wearable-based psychological stress detection and classification
• Affective computing and emotion recognition using physiological signals
• Multimodal sensor fusion for health state inference
• Benchmark testing of machine learning models on real IoT wearable data

Source & Attribution

WESAD was introduced by Schmidt et al. at the ACM International Conference on Multimodal Interaction (ICMI 2018). It is hosted on the UCI Machine Learning Repository and widely accessible for academic research.