/EEG-Motor-Imagery-Classification-CNNs-TensorFlow

EEG Motor Imagery Tasks Classification (by Channels) via Convolutional Neural Networks (CNNs) based on TensorFlow

Primary LanguagePython

EEG Motor Imagery Signals (Tasks) Classification via Convolutional Neural Networks (CNN)

Author: Shuyue Jia, Northeast Electric Power University, China.

Date: December of 2018

Download Paper

A novel approach of decoding EEG four-class motor imagery tasks via scout ESI and CNN

NOTICE: The method in our paper is EEG source imaging (ESI) + Morlet wavelet joint time-frequency analysis (JTFA) + Convolutional Neural Networks (CNNs). The raw data has been processed using the Matlab Toolkit Brainstorm. My job is using CNNs to classify the EEG data after the ESI + JTFA process. The Dataset (.mat Files) preprocessed via the ESI + JTFA process can be found via the Shared Google Drive. The corresponding preprocessed .Excel files can be downloaded from the Shared Google Drive.

Meanwhile, the codes in this repository are based on the raw EEG data without the ESI and JTFA process, and can also achieve a good result. The main CNNs Tensorflow framework codes in the "MI_Proposed_CNNs_Architecture.py" are the same for both of the work.


Project2

Installation and Usage

  1. Python file: PhysioNet_MI_Dataset/MIND_Get_EDF.py

    --- download all the EEG Motor Movement/Imagery Dataset .edf files from here!

    (Under Any Python Environment) $ python MIND_Get_EDF.py
    
  2. Python file: Read_Raw_Data_Save_Into_Matlab_Files.py

    --- Read the edf Raw data of different channels and save them into matlab .m files

    --- At this stage, the Python file must be processed under a Python 2 environment (I recommend to use Python 2.7 version).

    (Under Python 2.7 Environment) $ python Read_Raw_Data_Save_Into_Matlab_Files.py
    
  3. Matlab file: Saved_Matlab_Data/Preprocessing_Raw_Data.m

    --- Pre-process the dataset (Data Normalization mainly) and save matlab .m files into Excel .xlsx Files

  4. Python file: MI_Proposed_CNNs_Architecture.py

    --- the proposed CNNs architecture

    --- based on TensorFlow 1.12.0 with CUDA 9.0 or TensorFlow 1.13.1 with CUDA 10.0

    --- The trained results are saved in the Tensorboard

    --- Open the Tensorboard and save the results into Excel .csv files

    --- Draw the graphs using Matlab or Origin

    (Under Python 3.6 Environment) $ python MI_Proposed_CNNs_Architecture.py
    

Structure of the code

At the root of the project, you will see:

├── PhysioNet_MI_Dataset
|  └── MIND_Get_EDF.py
├── Read_Raw_Data_Save_Into_Matlab_Files.py
├── Saved_Matlab_Data
|  └── Preprocessing_Raw_Data.m
├── MI_Proposed_CNNs_Architecture.py
├── electrode_positions.txt

Citation

If you find our work useful in your research, please consider citing in your publications. We provide a BibTeX entry below.

@article{hou2019novel,  
    year = 2020,  
    month = {feb},  
    publisher = {{IOP} Publishing},  
    volume = {17},  
    number = {1},  
    pages = {016048},  
    author = {Yimin Hou and Lu Zhou and Shuyue Jia and Xiangmin Lun},  
    title = {A novel approach of decoding {EEG} four-class motor imagery tasks via scout {ESI} and {CNN}},  
    journal = {Journal of Neural Engineering}  
}  

Acknowledgment

We are very grateful to Prof. Yimin Hou due to his friendly guidance, and the research paper would not have happened without him.