eeg-human-action-classification

This repository contains example code for classifying human actions (specifically, left or right hand movements) using real EEG data. Additionally, it includes tools for integrating EEG data with the Robot Operating System (ROS).

Getting Started

Prerequisites

• Python 3.X
• ROS 1 Noetic (optional for ROS tools)

Installation

1. Clone this repository

git clone https://github.com/yourusername/eeg-human-action-classification.git
cd eeg-human-action-classification

2. Install the required Python libraries

pip install -r requirements.txt

Usage

1. EEG action classification:

Download the datasets under data folder from the link. Run the script to process EEG data and classify hand movements:

python ./scripts/train.py

2. EEG UDP ROS node for Bittium NeurOne To run the EEG UDP ROS node for Bittium NeurOne, use the following command:

python ./ros_utils/nerone_udp_node.py

You need to set the following arguments based on your device settings:

• n_electrodes: The number of electrodes to read.

• packet_size: The number of data points per electrode.

• device: The Bittium NeurOne device ("exg" or "tesla").

• channel_type: The type of channel reading electric type ("ac" or "dc").

• udp_ip: The IP address of the device (this must match the device software and the receiving computer's network settings).

• udp_port: The port number of the device (this must match the device software and the receiving computer's network settings).

• Example command:

python ./ros_utils/nerone_udp_node.py --n_electrodes 32 --packet_size 1 --device "tesla" --channel_type "ac" --udp_ip "192.168.200.240" --udp_port 50000

License

This project is licensed under the Apache License Version 2.0. See the LICENSE file for details.

Acknowledgement

These tools were originally developed for the following work.

@article{Choi2024OnTF,
  title={On the Feasibility of EEG-based Motor Intention Detection for Real-Time Robot Assistive Control},
  author={Ho Jin Choi and Satyajeet Das and Shaoting Peng and Ruzena Bajcsy and Nadia Figueroa},
  journal={ArXiv},
  year={2024},
  volume={abs/2403.08149},
  url={https://api.semanticscholar.org/CorpusID:268379125}
}