WAUC Dataset
"WAUC: A Multi-Modal Database for Mental Workload Assessment Under Physical Activity"
- Paper (Frontiers in Neuroscience)
- Dataset Website
- Original Dataset (Kaggle)
- Cleaned Dataset by Session (Kaggle)
Overview
WAUC is a large multimodal database for mental Workload Assessment Under physical aCtivity. It uniquely addresses the challenge of monitoring cognitive load when the user is simultaneously physically active - a scenario directly relevant to many real-world deployments, including military operators, industrial workers, and physically active emergency responders.
Study Design
- 48 participants - the largest participant count across the datasets in this project
- Cognitive task: NASA MATB-II (Multi-Attribute Task Battery II) - a realistic multi-task cognitive battery originally developed to simulate aircraft monitoring. Participants simultaneously manage system monitoring, tracking, communications, and resource management.
- Three physical activity conditions designed to introduce motion artefacts and physiological confounds:
| Condition | Activity | Speed |
|---|---|---|
| Rest (control) | Seated | None |
| Stationary bike | Cycling | Varying speeds |
| Treadmill | Walking/running | Varying speeds |
This multi-condition design enables systematic study of how motion artefacts interact with cognitive signals, and how physically activity affects the physiological baseline for cognitive state prediction.
Cognitive Labels
| Instrument | Description | Timing |
|---|---|---|
| NASA-TLX | Post-section workload rating across all six dimensions | After each experimental section |
| Borg Fatigue Scale | Physical fatigue self-rating | After each physical activity section |
Sensor Suite
WAUC is the most comprehensively instrumented dataset in this project, with simultaneous recordings across six physiological modalities:
| Sensor | Signal | Sampling Rate | Channels |
|---|---|---|---|
| Neuroelectrics Enobio 8 | EEG | 500 Hz | 8 ch: AF8, FP2, FP1, AF7, T10, T9, P4, P3 |
| Empatica E4 | Blood Volume Pulse (BVP) | 64 Hz | 1 |
| Empatica E4 | Galvanic Skin Response (GSR) | 4 Hz | 1 |
| Empatica E4 | Skin temperature | 4 Hz | 1 |
| Empatica E4 | Accelerometry (3-axis) | 32 Hz | 3 |
| Bioharness 3 | ECG | 250 Hz | 1 |
| Bioharness 3 | Respiration rate | 25 Hz | 1 |
| Bioharness 3 | Accelerometry (3-axis) | 18 Hz | 3 |
Enobio 8 Electrode Coverage
The 8 electrodes (AF7, AF8, FP1, FP2, T9, T10, P3, P4) cover:
- Frontal poles (FP1, FP2) - prefrontal cortex; decision-making and working memory.
- Anterior frontal (AF7, AF8) - frontal lobe activity adjacent to PFC.
- Temporal (T9, T10) - memory and emotion.
- Parietal (P3, P4) - spatial processing, attention.
This placement overlaps significantly with the NASA-TLX stressor-electrode mappings for Mental Demand (frontal), Effort (frontal + parietal), and Frustration (frontal + temporal).
Data Quality Notes
| Subject | Issue |
|---|---|
| S20 | Bioharness 3 not present - no ECG or respiration data |
| S25 | Empatica E4 not present - no BVP, GSR, temperature, or wrist accelerometry |
| S28 | No subjective metrics collected - cannot be used for supervised fine-tuning |
Why WAUC is Valuable
| Property | Value for Brain FM |
|---|---|
| 48 participants | Largest participant count; maximises demographic diversity for fine-tuning |
| 6 simultaneous modalities | Gold standard for multimodal fusion and cross-modal alignment research |
| Physical activity conditions | Enables study and mitigation of motion artefacts in both EEG and PPG |
| ECG included | Primary source for ECG cross-modal alignment with EEG |
| NASA MATB-II task | Realistic multi-attribute monitoring task; high face validity |
Relevance to the Brain FM Project
WAUC serves multiple roles:
- EEG pre-training data - 48 participants of Enobio 8 EEG during cognitive tasks.
- Multimodal alignment - simultaneous EEG + BVP + ECG enables three-way cross-modal alignment.
- Motion artefact robustness testing - physical activity conditions test the BFM's resilience to motion-corrupted signals.
- Device agnosticism evaluation - Enobio 8 uses different electrode positions and sampling rate (500 Hz) from UNIVERSE's Muse S (256 Hz); combining them tests device agnosticism.