/EEG-Emotion-Recognition

This repository compares typical and advanced modeling approaches for EEG Emotion Recognition.

Primary LanguageJupyter Notebook

EEG Emotion Recognition (DEAP dataset)

This project is still on progress

Authors: myself and my Ph.D. student Mr. Akraradet, and Master students Ms. Pranissa, Ms. Chanapa, and Mr. Pongkorn.

This repository compares different modeling approaches ranging from traditional machine learning algorithms and deep learning algorithms, by using emotion recognition from the benchmark DEAP dataset as the case study.

My intention is that there are just so many researches out there about the DEAP dataset but they can be hardly compared. Consequently, as a EEG researcher, it is almost impossible to know what architectural decisions should I make. This is due to the fact that some paper either did not provide the codebase hence not reproducible, or did not clearly specify the hyperparameters used, or just simply due to the obvious fact that even two papers using the same model cannot be directly compared because of differences in hardware and hyperparameters used.

Thus I want to make a controlled comparision of typical EEG models to create a clear understanding what works or what does not.

My another intention is that this codebase can be used by my Master and Ph.D. students as the getting started kit for their EEG research, since it mostly covers most of the typical EEG models.

Note: Before using the tutorials, please create a folder "data" and download preprocessed DEAP dataset and put s01.dat,...,s32.dat inside this "data" folder. The data folder will be in the same directory as the tutorial. Also create an empty "models" folder as well.

Python libraries:

  1. Python MNE
  2. PyTorch
  3. NumPy
  4. Sckit-Learn
  5. SciPy

Docker Prerequisite

  1. Docker
  2. Docker-Compose

How to use

The docker is designed to use with Visual Studio Code with Docker Extension. This way we can attach visual code to the docker environment.

Once you compose the service, go to the docker tab and find eeg-emotion. Right click and select Attach Visual Studio Code. Open the /root/projects/ and have fun coding.

There are 2 types of docker-compose. CPU only and GPU

  • CPU
docker-compose -f docker-compose-cpu.yml up --build -d
  • GPU
docker-compose -f docker-compose-gpu.yml up --build -d

Result DEAP + SVM with mulitple variables

Segment First

60s 30s 20s 12s 5s 4s 3s 2s 1s avg
DE 0.636±0.040 0.647±0.038 0.653±0.029 0.659±0.029 0.664±0.019 0.669±0.018 0.664±0.016 0.675±0.014 0.677±0.010 0.660±0.024
AS (DASM) 0.633±0.038 0.656±0.036 0.651±0.030 0.652±0.026 0.648±0.018 0.653±0.016 0.645±0.013 0.650±0.012 0.645±0.008 0.648±0.022
AS (RASM) 0.621±0.050 0.636±0.041 0.628±0.027 0.625±0.021 0.617±0.008 0.615±0.009 0.614±0.005 0.615±0.004 0.614±0.003 0.621±0.019
AS (DCAU) 0.641±0.046 0.651±0.038 0.654±0.032 0.654±0.028 0.648±0.017 0.652±0.017 0.643±0.014 0.648±0.012 0.644±0.008 0.648±0.024
CN ($PCC_{time}$) 0.625±0.024 0.633±0.025 0.635±0.178 0.639±0.019 0.641±0.014 0.676±0.016 0.645±0.011 0.647±0.010 0.634±0.010 0.642±0.034
CN ($PCC_{freq}$) 0.636±0.042 0.648±0.046 0.648±0.037 0.655±0.032 0.662±0.021 0.642±0.022 0.659±0.019 0.662±0.014 0.657±0.011 0.652±0.027
CN (PLV) 0.654±0.040 0.668±0.043 0.676±0.038 0.686±0.032 0.693±0.021 0.659±0.019 0.696±0.018 0.698±0.015 0.688±0.011 0.680±0.026
CN (PLI) 0.615±0.048 0.623±0.039 0.619±0.032 0.618±0.026 0.615±0.016 0.696±0.021 0.613±0.013 0.613±0.011 0.611±0.008 0.625±0.024
CSP 0.840±0.040 0.790±0.041 0.787±0.036 0.790±0.023 0.753±0.020 0.750±0.016 0.745±0.016 0.732±0.014 0.763±0.009 0.772±0.024
avg 0.656±0.041 0.661±0.039 0.661±0.049 0.664±0.026 0.660±0.017 0.668±0.017 0.658±0.014 0.660±0.012 0.659±0.009

Split First

60s 30s 20s 12s 5s 4s 3s 2s 1s avg
DE 0.580±0.134 0.588±0.130 0.593±0.120 0.590±0.110 0.589±0.105 0.595±0.104 0.587±0.103 0.594±0.101 0.591±0.096 0.590±0.111
AS (DASM) 0.575±0.135 0.584±0.127 0.587±0.125 0.586±0.115 0.580±0.106 0.584±0.107 0.580±0.103 0.581±0.103 0.578±0.990 0.582±0.212
AS (RASM) 0.564±0.130 0.578±0.125 0.568±0.123 0.567±0.119 0.552±0.123 0.552±0.122 0.553±0.127 0.552±0.124 0.553±0.127 0.560±0.124
AS (DCAU) 0.583±0.134 0.592±0.126 0.591±0.121 0.559±0.113 0.584±0.104 0.585±0.102 0.580±0.102 0.582±0.102 0.578±0.100 0.582±0.112
CN ($PCC_{time}$) 0.564±0.135 0.570±0.128 0.572±0.124 0.573±0.119 0.574±0.116 0.574±0.114 0.573±0.116 0.575±0.114 0.574±0.110 0.572±0.120
CN ($PCC_{freq}$) 0.578±0.130 0.585±0.122 0.585±0.116 0.585±0.110 0.585±0.106 0.581±0.102 0.581±0.100 0.582±0.099 0.577±0.096 0.582±0.109
CN (PLV) 0.591±0.130 0.605±0.119 0.609±0.114 0.611±0.108 0.607±0.102 0.604±0.100 0.603±0.098 0.602±0.096 0.596±0.092 0.603±0.107
CN (PLI) 0.564±0.124 0.562±0.117 0.559±0.113 0.560±0.111 0.554±0.107 0.555±0.106 0.553±0.104 0.554±0.104 0.554±0.103 0.557±0.110
CSP 0.830±0.074 0.758±0.085 0.746±0.080 0.731±0.079 0.704±0.079 0.703±0.080 0.700±0.081 0.681±0.080 0.720±0.074 0.730±0.079
avg 0.603±0.125 0.602±0.120 0.601±0.115 0.596±0.109 0.592±0.105 0.593±0.104 0.590±0.104 0.589±0.103 0.591±0.199