Most people face stress at some point in their lives, whether it be physical or emotional. Stress has a negative impact on a person's health. Since, research on stress is still in its infancy, and over the past 10 years, much focus has been placed on the identification and classification of stress. We presented an end-to-end solution for detection of stress from EEG signals collected from an OpenBCI Ganglion EEG Headset. We examined an LSTM and GRU model for classifying stress. We showcase a simple music recommendation system using the Spotify API to play music based on the user's current mood. We also present two architectures to scale this application for large scale use. Please find the presentation for this project here.
The following components were required for the setup
- OpenBCI Ganglion 4-channel EEG Headset
- Electrode wires
- Gel electrode stickers
- Duracell AA Batteries
2 electrodes of OpenBCI Ganglion were used as reference and ground placed on both earlobes. The other 4 electrodes were placed on location TP9, AF7, AF8, TP10, as per the 10-20 system of electrode placement.
- Step 1: Collect EEG Data by placing the electrodes in the locations TP9, AF7, AF8, TP10.
- Step 2: Pre-process the data using this library.
- Step 3: Train the model on a publically available kaggle dataset that resembles the recorded data.
- Step 4: Infer on the incoming raw data. The incoming data is split into 1-min chunks to detect the emotion.
- https://github.com/jordan-bird/eeg-feature-generation
- https://brainflow.org/
- https://streamlit.io/
- https://www.kaggle.com/datasets/birdy654/eeg-brainwave-dataset-feeling-emotions
Vijayasri Iyer & Madhumithaa V