This repository contains the source code for the system presented as a demo at the IEEE Engineering in Medicine and Biology Society (EMBS).
Title: Brain-Controlled Robot for Assisting Basic Upper Limb Tasks
Link: Read the full paper on ResearchGate
- Acquisition: The Muse 2 headband measures brain activity (Alpha, Beta, Gamma, Delta, Sigma waves) via 4 electrodes.
- Streaming: Data is streamed wirelessly to the Laptop via UDP.
- Inference: A Python script processes the live EEG feed using a trained classifier to detect mental states (e.g., Focus, Relax, Blink artifacts).
- Actuation: The predicted class is converted into a motor command and sent to the Raspberry Pi via TCP, which drives the servo motors.
| File Name | Description |
|---|---|
Collect EEG-data.py |
Data Acquisition: Listens to the UDP stream from the Muse sensor, labels the data, and saves it for training. |
Models/ |
Saved Models: Directory containing the pre-trained TensorFlow .h5 models. |
Test Classifier Model.py |
Real-Time Inference: Loads the trained model, classifies live EEG data, and sends commands to the robot. |
Muse-Robot-Control.py |
Heuristic Control: A simpler control method using raw artifacts (Jaw Clenches / Blinks) instead of the ML model. |
Rasp-Arm-tcpPI.py |
Server (Robot Side): Runs on the Raspberry Pi. Listens for TCP commands and drives the motors. |
Gui-tcpPC.py |
Testing Tool: A PC-side GUI to manually test the TCP connection and robot movements. |
- EEG Headset: Muse 2 (or compatible OSC-streaming headband).
- Robot: 4-DOF Robotic Arm (Servos + Driver Board).
- Controller: Raspberry Pi (3 or 4 recommended).
- Computer: Laptop with Python environment for heavy processing.