WAUC: A Multi-Modal Database for Mental Workload Assessment Under Physical Activity

This repository provides codes and documentation for the new release of the WAUC dataset. The following changes/additions have been made for this new release:

• Data has been reformatted into easy to read csv format.
• The csv files have data separated and classified into sessions (1 to 6) along with baselines (eyes closed/still and movement only)
• A data integrity section has been provided which provides information regarding missing sensors or sessions
• Demographc info has been provided in demographics.csv file including age, and sex.
• More detail has been provided regarding the data structure and format for easier access and analysis

Accessing the Dataset

A sample dataset for one subject is provided at the following link

Citation

If you use this dataset in your research, please cite it as follows:

@article{albuquerque2020wauc,
  title={Wauc: a multi-modal database for mental workload assessment under physical activity},
  author={Albuquerque, Isabela and Tiwari, Abhishek and Parent, Mark and Cassani, Raymundo and Gagnon, Jean-Fran{\c{c}}ois and Lafond, Daniel and Tremblay, S{\'e}bastien and Falk, Tiago H},
  journal={Frontiers in Neuroscience},
  volume={14},
  pages={549524},
  year={2020},
  publisher={Frontiers Media SA}
}

Data Integrity

There are few instances of missing sensors and data. These include:

Bioharness 3

The instances of missing data for BH3 include:

Subject	Description
1004	Session #5 missing (high physical activity)
1013	No baseline markers for session #1
1019	Session #3 missing (high physical activity)
1020	No BH3 data
1032	Session #2 missing (high physical activity)
1035	Session #1 missing

Empatica E4

The instances of missing data for E4 include:

Subject	Description
1001	No E4 data
1013	No baseline markers for session #1
1025	No E4 data
1028	No E4 data
1038	No E4 data

Enobio EEG Cap

For the Enobio EEG cap, only baseline markers for Subject 1013 for session 1 are missing

Ratings

Subject 1028 has no ratings or session ground truth info avalible

Data Structure and Format

This section covers how the dataset is organized and the structure of the csv files

Datset Directory Structure

WAUC_Dataset
├── 1001/
├── 1002/
│   ├── raw/
│   │   ├── {sensor1}_{signal1}.csv
│   │   ├── {sensor1}_{signal2}.csv
│   │   ├── {sensor2}_{signal1}.csv
│   │   ├── {sensor2}_{signal2}.csv
│   │   └── ... 
│   ├── processed/
│   └── features/
├── 1003/
├── ...
├── ...
├── 1047/
├── 1048/
├── subjective_ratings_with_labels.csv
└── demographics.csv

Each subject folder has raw, processed, and features subfolders. For now, only the raw folder is populated and includes the raw sensor information collected named as {sensor}_{signal}.csv, the sensors and corresponding signals are shown in table below:

Sensor	Signals
bh3	acc, ecg, br
e4	acc, ppg, temp, gsr
enobio	eeg

The subjective ratings including NASA-TLX and percieved exertion scale along with ground truth of MATB-II (column: mwl) and physical activity (column: pwl) is provided in subjective_ratings_with_labels.csv. These values can be mapped to their corresponding physiological information using the session number columns (column: session_no)

The demographics.csv file includes age, sex (M or F), weight (kg), and height (cm) along with the activity type (Treadmill or Bike) performed by the participants.

Signal CSV File Structure

All signal csv files contain the following columns:

Columns	Description
unix	Unix time of experiment
time	Time (in s) from the start of recording
fs	Sampling Frequency of the signal
info	baseline-1 = eye closed + still, baseline-2 = physical activity only, session = main recording session
session_no	Session number (same as subjective ratings file)