This project was tested with: Anaconda Distribution 2022.05 For Python 3.10 on Windows 11
The goal of this project is to experiment with the WESAD multimodal dataset to become famliar with Python for deep learning. The project uses different ML models to predict when an individual reaches a state of stress or increased stress.
WESAD, a multimodal WEarable Stress and Affect Detection dataset, is a collection of physiological signal data recorded from 12 subjects in which the experimenters induced a stressful state using the TSST (Tier Social Stress Test). Multiple physiological parameters in addition to acceleration information were captured during the experiments. They attached two diferent devices to each individual; one wrist worn and one chest worn and collected multiple modalities for each.
This project considers the chest worn device with a subset of the total number of modalities.
WESAD dataset ~17 GB of raw data. Download link is here.
Download and extract WESAD data set so that each subject has a folder (SX, where X = subject ID). Each subject folder contains the following files:
- SX_readme.txt: contains information about the subject (SX) and information about data collection and data quality (if applicable)
- SX_quest.csv: contains all relevant information to obtain ground truth, including the protocol schedule for SX and answers to the self-report questionnaires
- SX_respiban.txt: contains data from the RespiBAN device
- SX_E4_Data.zip: contains data from the Empatica E4 device
- SX.pkl: contains synchronised data and labels
- 17 subjects participated in the study. However, due to a sensor malfunction, data from two subjects (S1 and S12) had to be discarded.
- WESAD_all_models.ipynb contains all visualization techniques applied to the dataset
- WESAD_visualizationa.ipynb consists of preprocessing along with different ML models such as LDA, QDA, Decision Trees, KNN along with techniques such as k-Fold Cross Validation along with their analysis and optimization strategies employed.
- Python3
- pandas
- matplotlib
- numpy
- sklearn
- graphviz
- seaborn
- xgboost
- statsmodel.api
- pickle
- ipython and jupyter for jupyter notebook
To run the notebooks and the python module:
- (1) Set a path to the git project
module = os.path.abspath('C:/github/stress-affect-wesad/src/main') - (2) Set a path to the WESAD dataset
ROOT_PATH = 'C:/Users/dollj/OneDrive/Desktop/datasets/WESAD' - (3) Make sure to point DataManager.py and the Notebooks to the data source. See below example.
- (4) Run Demo.py to prepare, create, and evaluate an LSTM NN with one epoch and also load and evalute an LSTM NN with 5 epochs.
File Structure
/
demo.sh - driver that sets up, builds, trains, and tests model
readme.md
WESAD_all_models.ipynb - does not use Keras framework
WESAD_visualization.ipynb - does not use Keras framework
references/ - WESAD dataset and paper information
src/
src/main - Python3 modules
DataManager.py
Demo.py
src/notebooks - jupyter ipython notebooks
exploring-the-dataset.ipynb
feature-exploration.ipynb
feature-exploration-continued.ipynb
model-training.ipynb
demo
src/models - Directory for Keras model data files
to do - add notes on development process and WESAD_all_models.ipynb performance using Windows
Performance using the LSTM based network architecture with one hidden layer has performed with an accuracy of ~ 97.4% on the validation set using 5 epochs.
Windows: Learning rate = 0.05 batch size = 2 With just one epoch, the model has results between ~80% and ~92% for accuracy on the validation data. Each epoch of training takes approximately 70 seconds without GPU acceleration. At 5 epochs, the model outperforms the WESAD quoated results for both accuracy and F1 using less modalities and less features.
- Add encoder-decoder model with attention layer
- Add notebook with LDA, QDA, Decision Trees, KNN, SVM, PCA, and k-fold cross validation
- Update directions in README.md with notes on development process and performance using Windows
Also, many thanks to the below mentioned authors for their past work:
-> https://github.com/BradySheehan/wesad_experiments