Neural Expressions

Experimentations with Brain-to-Sculpture communication and EEG actuated robotic sculptures

Funding

Project funded by DeepCA (Norwegian Research Council, grant agreement 286558) and FeLT (Norwegian Artistic Research Programme).

Directory structure

• README.md This file
• notebook.ipynb Showcase and notebook over the project
• code
  • requirements.txt Python libraries needed, python 3.8 used
  • bci
    • cyton_get_data.py Simple test with brainflow library for streaming EEG data to computer and getting EEG board information
    • lslStreamTest.py Tests for LSL stream, forked from https://github.com/OpenBCI/OpenBCI_GUI/tree/master/Networking-Test-Kit
    • brainflow Brainflow library. Not needed with 'pip install brainflow'
  • robot_control
    • osc_send.py Testing OSC messaging from computer via clt
    • osc_receive.py Testing reception of OSC messages
    • redboard.py Redboard library. Not needed with 'pip install redboard'
    • redboard_servo_control_test.py Stand-alone cli script for testing motor control
    • redboard_servo_control.py Simple functions for redboard library for moving a 2-wheeled robot car based on received OSC messages
    • rpimotorlib_motor_control.py Unfinished script for motor control with RpiMotorLib library
    • rpimotorlib_stepper_control.py Unfinished script for stepper control with RpiMotorLib library
  • machine_learning
    • get_npydata.py Saves LSL stream as 3-dimensional numpy array ( [stream][channels][amplitude] )
    • CytonUserSettings.json Just for reference: Settings used in OpenBCI GUI
    • data Numpy arrays with FFT stream from EEG recordings saved as binary in [Unix time].npy files
      • left
      • none
      • right
    • testing_and_making_data.py Saves LSL stream to /data, inferes stream towards model, and sends result over OSC. Forked from https://github.com/Sentdex/BCI/
    • accuracies.csv Record over accuracy of models, generated by testing_and_making_data.py
    • traning.py Trains data on 1D Convolutional network, saves models for each epoch in /new_models
    • new_models
    • validation_data Validation data excluded from /data
      • left
      • none
      • right
    • analysis.py Tests model by comparing predicted action vs action thought, and produces a confusion matrix
    • models Stored best-of models
• img Images used in jupyter notebook
• LICENSE.md Copyright information

License

MIT License Copyright © 2021 Daniel Slåttnes